OpenAI

ChatGPT’s voice mode gets integrated directly into the main chat interface

November 25, 2025 - The voice feature of ChatGPT is now more seamlessly integrated — you no longer need to jump to a separate interface just to talk with the chatbot. OpenAI announced on Tuesday that it is updating the interface so users can engage with ChatGPT Voice right within their ongoing chats.

From now on, you’ll be able to converse with the chatbot and simultaneously view its responses — including any shared images — without leaving the main chat. Previously, ChatGPT Voice occupied a different screen, represented by an animated blue circle, complete with its own controls: a mute button, an option to record video live, and an “X” to exit back to text mode. During those voice sessions, replies were only audible — never displayed as text in the main chat — meaning if you missed something, you’d need to exit voice mode to review the response on-screen.

With the update, responses will appear directly in the chat as the AI speaks, letting you both hear and read them. You’ll also be able to scroll through earlier messages and view visual elements like images or maps in real time.

This integration aims to make interacting with the chatbot more fluid, letting users transition between speaking and reading within the same conversation. Still, to return to purely text-based interaction, you’ll need to tap “end” to terminate the voice conversation.

The new voice mode is now the default and is being rolled out to all users across web and mobile platforms. Users who preferred the older, separate voice interface haven’t been forgotten: OpenAI says you can revert to it by navigating to “Voice Mode” under “Settings,” where a new “Separate mode” toggle is available.

Source: https://techcrunch.com/2025/11/25/chatgpts-voice-mode-is-no-longer-a-separate-interface/

OpenAI and Target Partner to Launch AI-Driven Retail Experiences via ChatGPT

November 19, 2025 - OpenAI and Target have expanded their partnership with the launch of a new Target app integrated into ChatGPT, designed to deliver a curated and conversational shopping experience. This move reflects Target’s broader strategy of embedding artificial intelligence across its operations to enhance both customer interactions and internal productivity.

Through the Target app in ChatGPT, customers will gain access to personalized recommendations, build multi-item shopping baskets, and complete purchases using Drive Up, Order Pickup, or shipping. The app, launching in beta next week, allows users to ask for shopping ideas—such as planning a family holiday movie night—and instantly receive tailored suggestions like snacks, blankets, and candles, all within a conversational interface.

“As AI transformation accelerates within enterprises, Target exemplifies how to implement it with ambition and speed,” said Fidji Simo, CEO of Applications at OpenAI. “We’re excited to support Target in creating useful and joyful experiences for their customers and employees.”

Target continues to rely on OpenAI’s APIs and ChatGPT Enterprise to empower its teams and streamline guest experiences. The enterprise-grade platform, which safely integrates proprietary data, has already been deployed across Target headquarters, benefiting over 18,000 employees. It enables faster workflows, reduces repetitive tasks, and supports creative problem-solving.

According to Prat Vemana, Executive Vice President and Chief Information and Product Officer at Target, “At Target, everything starts with the guest, and that means meeting them wherever they are—including emerging spaces like ChatGPT, where millions of consumers visit. Our goal is simple: make every interaction feel as natural, helpful, and inspiring as chatting with a friend.”

AI is also embedded into Target’s internal and external operations. Tools like Agent Assist and Store Companion help service teams resolve customer issues swiftly, including price matching and returns. The Gift Finder and Shopping Assistant offer AI-driven browsing and recommendation capabilities, while the JOY solution—trained on over 3,000 FAQs—instantly supports vendors and guests.

Future enhancements to the ChatGPT experience will include Target Circle integration and same-day delivery options.

This expanded partnership signifies a shift in the retail landscape, where enterprises are not just adopting AI tools but integrating them into the core of their business models. With over one million business customers, OpenAI’s platform supports wide-ranging applications across industries, making it a key player in driving the AI-led transformation of retail.

Source: https://openai.com/index/target-partnership/

OpenAI Launches Group Chat Feature in ChatGPT for Collaborative AI Conversations

November 14, 2025 - OpenAI has launched a new group chat feature for ChatGPT, marking the platform’s first shared conversation experience. The update allows up to 20 users to collaborate in a single chat, with real-time AI support. Initially available as a pilot in Japan, New Zealand, South Korea, and Taiwan, the feature will expand globally following user feedback.

Describing the rollout, OpenAI stated, “With group chats, you can bring friends, family, or coworkers into a shared space to plan, make decisions, or work through ideas together.” The feature integrates collaboration tools directly into the ChatGPT interface, putting it in line with offerings from other AI platforms.

To initiate a group chat, users can tap the new people icon in the app. ChatGPT generates a shareable link that allows any account holder—free or paid—to join the conversation. Each group can host up to 20 members, and existing one-on-one threads can be duplicated into new group chats to preserve private messages.

Participants create a profile with a name, username, and optional photo. Chats are stored in a dedicated sidebar, and can be used for a range of purposes from trip planning to work collaborations. Custom instructions can be set per group, enabling the AI to adapt its tone for different settings.

OpenAI noted that ChatGPT responds differently in group environments. Instead of replying to every message, it monitors the discussion and interjects at suitable moments. Users can also tag the AI for input. The assistant now supports emoji reactions, adding a layer of interaction.

The platform’s group chat feature includes full access to tools like web search, image uploads, file sharing, voice dictation, and image generation. It runs on GPT-5.1 Auto, which dynamically selects the best model available to each user based on their subscription level, ensuring tailored responses within the same chat.

Creators of a group can manage participants, while anyone can rename the chat or mute notifications. Privacy safeguards include the exclusion of memory from group chats and the activation of sensitive content filters when minors are present. Parents can also choose to disable group chats entirely.

Though still in its pilot phase, the feature is expected to become a key part of ChatGPT’s offering as it moves toward global release.

Source: https://www.indiatoday.in/technology/news/story/openai-launches-group-chats-in-chatgpt-sam-altman-gpt-5-2819513-2025-11-14

OpenAI Explores Consumer Health Tools, Eyes AI-Powered Personal Health Assistant

November 10, 2025 - OpenAI is considering a move into consumer healthcare, with potential plans to develop generative AI-powered personal health tools, according to a report from Business Insider published Monday. The initiative marks a strategic shift for the ChatGPT creator as it looks to expand beyond its core AI offerings.

While OpenAI declined to comment on the report, the company’s recent hiring activity suggests growing interest in healthcare. In June, OpenAI appointed Nate Gross, co-founder of physician networking platform Doximity, as head of healthcare strategy. In August, former Instagram executive Ashley Alexander joined as vice president of health products.

At the HLTH conference in October, Gross revealed that ChatGPT sees roughly 800 million weekly active users, with a significant portion seeking medical-related information or advice.

The move positions OpenAI alongside tech giants like Google, Amazon, and Microsoft, all of which have previously made attempts to help consumers manage their health data—with mixed results. Google shuttered its health record service in 2011 due to limited uptake. Amazon discontinued its Halo fitness tracker business in 2023, and Microsoft's HealthVault platform failed to gain broad consumer adoption.

OpenAI’s potential entry into health-focused AI tools could reshape how individuals interact with personal medical information, though the company has yet to publicly confirm any products or launch timelines.

Source: https://www.reuters.com/business/healthcare-pharmaceuticals/openai-considers-consumer-health-tools-push-beyond-core-ai-offerings-business-2025-11-10/

ChatGPT Barred from Offering Medical, Legal, or Financial Advice Amid Liability Concerns

November 4, 2025 - OpenAI has reclassified ChatGPT as an “educational tool,” effectively restricting the AI system from providing specific medical, legal, or financial advice. The policy shift, attributed to increasing regulatory scrutiny and potential liability issues, was first reported by Nexta.

Under the new guidelines, ChatGPT users are prohibited from using the model for consultations that would typically require certified professionals. This includes areas such as healthcare, legal services, financial planning, housing, education, immigration, and employment—unless overseen by a qualified human expert.

While the AI will continue to offer general information—for example, explaining what a will is or how tax brackets function—it will no longer deliver personalized advice or professional-grade guidance. The updated terms also bar the use of ChatGPT for AI-assisted facial recognition without consent and explicitly prohibit actions that may lead to academic misconduct.

The company states the move is aimed at “enhancing user safety and preventing potential harm” resulting from misuse of the model in high-stakes scenarios.

ChatGPT is now limited to explaining principles, outlining general frameworks, and referring users to licensed professionals. It is no longer permitted to name specific medications or dosages, produce legal document templates, or provide investment strategies or buy/sell recommendations.

Attempts to bypass these limitations using hypothetical queries are reportedly blocked by updated safety filters. This change follows growing public debate over AI’s role in offering expert-level advice, especially in healthcare contexts.

Notably, conversations with ChatGPT do not benefit from doctor–patient or attorney–client confidentiality protections. As such, user interactions may be subject to legal discovery and could be subpoenaed in court.

In tandem with these restrictions, OpenAI has also rolled out new safety features to support users experiencing mental health issues, including psychosis, mania, self-harm, and suicidal ideation, as well as those displaying emotional dependency on AI.

Despite its utility in simplifying concepts, summarizing information, and ideation, OpenAI emphasizes that ChatGPT cannot substitute for licensed professionals. It lacks the capacity to interpret non-verbal cues, provide empathetic care, or guarantee user safety.

Users facing crises are advised to contact qualified professionals or emergency services, such as dialing 988 in the U.S., rather than rely on AI systems. Similarly, legal and financial decisions should be handled by certified advisors, as ChatGPT cannot assess individual circumstances, risk profiles, or jurisdictional regulations.

Using the model to draft legal or financial documents carries tangible risks, including potential errors or invalid content. The system is also unsuited for real-time emergencies, lacking the ability to detect hazards like gas leaks or contact authorities. While ChatGPT can reference web data, it does not provide continuous monitoring and may include outdated or inaccurate information.

Finally, users are warned against inputting sensitive or confidential data—such as financial statements, medical records, or private contracts—since data storage and access controls are not guaranteed under the current model.

Source: https://www.financialexpress.com/life/technology-openai-pulls-back-chatgpt-will-no-longer-give-medical-legal-or-financial-advice-over-liability-fears-4030898/

OpenAI Debuts Aardvark: An Autonomous AI Agent for Security Vulnerability Detection

October 31, 2025 - OpenAI has launched Aardvark, a new AI-powered agent built to autonomously detect and remediate security vulnerabilities in software codebases. Running on GPT-5, Aardvark is currently in private beta, with select partners helping fine-tune its accuracy, validation mechanisms, and reporting interface.

"Aardvark represents a breakthrough in AI and security research—an autonomous agent that can help developers and security teams discover and fix security vulnerabilities at scale," OpenAI stated during the announcement.

This tool actively scans code repositories, identifies and verifies vulnerabilities, evaluates their exploitability, and suggests appropriate patches. Unlike conventional techniques like fuzzing or software composition analysis, Aardvark leverages large language model (LLM) reasoning to interpret source code, detect issues, and auto-generate solutions.

Aardvark's operation involves multiple phases: analyzing entire code repositories to build a threat model, reviewing new commits for flaws, testing exploitability in isolated environments, and producing patches via Codex, intended for human assessment and implementation.

In internal trials, Aardvark successfully flagged 92% of both known and artificially inserted vulnerabilities within benchmark datasets. The tool has also been utilized across OpenAI's infrastructure and those of early collaborators, uncovering "meaningful vulnerabilities" that have reinforced security frameworks.

Additionally, OpenAI disclosed that Aardvark has been employed in various open-source projects, where it has uncovered and responsibly disclosed several security issues. Notably, ten of these have received official Common Vulnerabilities and Exposures (CVE) designations.

"As beneficiaries of decades of open research and responsible disclosure, we’re committed to giving back—contributing tools and findings that make the digital ecosystem safer for everyone," OpenAI emphasized. The organization also unveiled plans to offer complimentary scans for select non-commercial open-source codebases.

To support responsible coordination, OpenAI has revised its disclosure policies to encourage collaborative remediation efforts and realistic timelines. "We anticipate tools like Aardvark will result in the discovery of increasing numbers of bugs, and want to sustainably collaborate to achieve long-term resilience," the company added.

The launch coincides with growing concerns over software vulnerabilities. In 2024 alone, over 40,000 CVEs were recorded. OpenAI highlighted that approximately 1.2% of all code commits result in bugs.

With Aardvark, OpenAI aims to shift cybersecurity toward a proactive, defender-focused paradigm, offering continuous defense as codebases evolve.

Source: https://analyticsindiamag.com/ai-news-updates/openai-launches-aardvark-an-ai-agent-for-automated-security-research/

OpenAI to Provide Free ChatGPT Go Access for One Year to Users in India

October 27, 2025 - OpenAI has announced that users in India will receive free access to its ChatGPT Go plan for one year, starting November 4, as part of a limited-time promotional offer. The offer applies to all new users signing up during the promo period, as well as existing ChatGPT Go subscribers in the country.

ChatGPT Go, OpenAI’s most affordable paid subscription, was introduced in India in August at a price of under $5 per month. The plan has since expanded to Indonesia and 16 more countries across Asia earlier this month.

India, with more than 700 million smartphone users and over a billion internet connections, remains a strategic market for OpenAI. The company launched its New Delhi office in August and is currently hiring to expand its operations locally.

CEO Sam Altman previously identified India as OpenAI’s second-largest market after the U.S. However, monetizing the platform has proven challenging. In the 90 days leading up to August, the ChatGPT app saw over 29 million downloads in India but generated only $3.6 million in in-app purchases, based on data from Appfigures.

The ChatGPT Go plan offers significantly enhanced features compared to the free version, including 10 times more usage for text generation, image creation, and file uploads. It also comes with improved memory for delivering more personalized responses.

“Since initially launching ChatGPT Go in India a few months ago, the adoption and creativity we’ve seen from our users has been inspiring,” said Nick Turley, vice president and head of ChatGPT. “We’re excited to see the amazing things our users will build, learn, and achieve with these tools.”

OpenAI’s move follows similar efforts by competitors aiming to tap into India’s vast digital user base. Perplexity, for example, recently partnered with Airtel to provide free access to its Pro plan for 360 million users. Google also introduced a free one-year AI Pro subscription plan specifically for students in India.

OpenAI is scheduled to host its DevDay Exchange developer conference in Bengaluru on November 4, where India-specific product announcements for developers and enterprises are expected.

Source: https://techcrunch.com/2025/10/27/openai-offers-free-chatgpt-go-for-one-year-to-all-users-in-india/

OpenAI Taps Broadcom for Custom AI Chip Partnership Worth Up to $500B

October 14, 2025 - OpenAI has announced a major partnership with semiconductor giant Broadcom to supply 10 gigawatts of custom AI accelerator hardware. These specialized AI racks will begin deployment across OpenAI and partner data centers in 2026, with rollout continuing through 2029.

“By designing its own chips and systems, OpenAI can embed what it’s learned from developing frontier models and products directly into the hardware, unlocking new levels of capability and intelligence,” the company said in an official statement.

Although specific financial terms were not shared, the Financial Times reported that the deal could cost OpenAI between $350 billion and $500 billion.

This Broadcom alliance is the latest in a series of major infrastructure moves by OpenAI.

Just last week, the company revealed plans to procure an additional six gigawatts of AI chips from AMD in a deal valued in the tens of billions. In September, Nvidia announced a $100 billion investment in OpenAI, along with a letter of intent to provide another 10 gigawatts of GPU infrastructure.

Reports have also surfaced of a potential $300 billion cloud services agreement between OpenAI and Oracle, though neither party has officially confirmed the deal.

Source: https://techcrunch.com/2025/10/14/openai-and-broadcom-partner-on-ai-hardware/

OpenAI Accelerates Developer Focus with Powerful New API Models

October 6, 2025 - At its Dev Day on Monday, OpenAI introduced a suite of API enhancements, headlined by the debut of GPT‑5 Pro, the new Sora 2 video model, and a smaller, more affordable voice model.

These updates are part of a broader strategy to deepen developer engagement. OpenAI also revealed a tool for building AI agents and enabled app creation directly within ChatGPT.

GPT‑5 Pro is tailored for industries requiring robust reasoning and high accuracy, such as finance, law, and healthcare. “High accuracy and depth of reasoning” are essential in these sectors, said OpenAI CEO Sam Altman.

Recognizing the growing role of voice in AI interactions, OpenAI launched “gpt‑realtime mini”, a compact voice model that supports low-latency streaming. It’s 70% cheaper than the previous advanced voice model, yet maintains “the same voice quality and expressiveness,” according to the company.

Developers can now preview Sora 2 in the API. The model also powers the new Sora app, a TikTok-style platform for short, AI-generated videos. Users can generate clips featuring themselves, others, or custom prompts, then share them through an algorithmic feed.

“[Developers] now have access to the same model that powers Sora 2’s stunning video outputs right in your own app,” Altman said.

Sora 2 improves on its predecessor with enhanced realism, synchronized audio, and finer creative controls, from camera direction to stylized rendering.

“For example, you can take the iPhone view and prompt Sora to expand it into a sweeping, cinematic wide shot,” Altman explained. “But one of the most exciting things that we’ve been working on is how well this new model pairs sound with visuals, not just speech, but rich soundscapes, ambient audio, synchronized effects that are grounded in what you’re seeing.”

Positioned as a concept development tool, Sora 2 can be used for anything from ad mockups to product ideation. One example shared at Dev Day featured toy maker Mattel using the model to turn sketches into toy concepts—part of a broader collaboration to integrate generative AI into product design.

Source: https://techcrunch.com/2025/10/06/openai-ramps-up-developer-push-with-more-powerful-models-in-its-api/

OpenAI’s Sora soars to No. 3 on the US App Store

October 2, 2025 - OpenAI's latest app, Sora, designed for AI-generated videos, has surged in popularity, reaching the No. 3 position on the U.S. App Store shortly after launch. Despite its invite-only model and limited availability to users in the U.S. and Canada, Sora managed to go viral. According to app analytics firm Appfigures, the iOS version of the app recorded 56,000 downloads on launch day alone.

In total, Sora amassed 164,000 installs across September 30 and October 1, per Appfigures' estimates.

Sora's debut outperformed other notable AI app launches, such as Anthropic's Claude and Microsoft's Copilot, and matched the opening-day numbers of xAI's Grok. However, OpenAI's own ChatGPT and Google's Gemini both saw stronger day-one performance, with at least 80,000 downloads each.

Given that Sora remains invite-only, direct comparisons with publicly accessible apps might not be entirely equitable. Still, the impressive uptake reflects growing consumer interest in AI-powered video tools, especially those offering a social media-like experience.

Appfigures conducted a focused analysis to offer a fair comparison, narrowing the scope to U.S. and Canadian downloads due to varying geographic launch strategies across AI apps. For instance, ChatGPT initially launched in the U.S. only, Grok targeted users in the U.S., Australia, and India, and Claude had no stated geographic limitations.

Through this apples-to-apples comparison, Appfigures reported the following day-one download numbers:

ChatGPT: 81,000

Gemini: 80,000

Sora and Grok: 56,000 each

Claude: 21,000

Copilot: 7,000

By its second day, Sora climbed to No. 3 on the U.S. App Store's overall top chart. For context, ChatGPT hit No. 1 on day two, Grok reached No. 4, Gemini landed at No. 6, Copilot came in at No. 19, and Claude trailed at No. 78.

While some within OpenAI may prefer the company to tackle more profound technological challenges, the rapid adoption of Sora suggests significant consumer demand for AI-driven creative tools—even if that demand includes quirky deepfake videos, like OpenAI CEO Sam Altman supposedly asking, "Are my piggies enjoying their slop?"

Source: https://techcrunch.com/2025/10/02/openais-sora-soars-to-no-3-on-the-u-s-app-store/

"GPT-5 Smarter Than Me": Sam Altman Predicts AI Could Replace 40% of Jobs by 2030

September 29, 2025 - Artificial intelligence is advancing rapidly and could significantly reshape the global workforce by the end of this decade, potentially replacing up to 40% of current jobs, according to OpenAI CEO Sam Altman.

Speaking at the Axel Springer Global Reporters Network event in Berlin, Altman highlighted the speed at which AI is evolving. "Around 1 per cent of tasks currently performed by humans are already being handled by AI. Over the next few years, 30 to 40 per cent of tasks could be done by AI," he stated.

Altman noted that while job evolution over decades is not new, the pace of transformation is accelerating. "In 30 years, jobs change all the time. If you think about the jobs we did three decades ago, many of them may no longer exist today or new jobs that were kind of difficult to imagine years ago are now commonplace today."

Referencing a statistic that suggests half of all jobs typically change every 75 years, Altman remarked that this shift is now occurring much faster. He added, "GPT-5 is already smarter than me and many others."

Although he sees this AI progress as a milestone in science and technology, Altman warned of the potential downsides. While AI could unlock innovations and solutions previously beyond human capability, it might also result in widespread job displacement. "By the end of this decade, if we don't have extraordinarily capable models that do things humans cannot, I'd be very surprised," he said.

In a recent appearance on The Tucker Carlson Show, Altman named programming and customer service as among the first sectors vulnerable to automation. He elaborated on the kinds of roles AI might replace versus those that would remain uniquely human.

"A job that I feel way less certain about what the future looks like for is computer programmers. A job that I'm confident will not be that impacted is, like, nurses," he commented.

Altman’s remarks add to the growing conversation around the transformative impact of AI on employment and the global economy.

Source: https://www.ndtv.com/world-news/gpt-5-smarter-than-me-sam-altman-warns-ai-could-replace-40-per-cent-jobs-by-2030-9362761

OpenAI Plans ChatGPT-Powered Hardware Device, Taps Apple Supply Chain for 2026–2027 Launch

September 21, 2025 -  OpenAI is gearing up to release its first hardware device between late 2026 and early 2027, according to a new report. The AI firm has begun recruiting hardware talent from Apple and is reportedly working with Luxshare, a key Apple supplier, to manufacture its upcoming product.

Earlier this year, OpenAI partnered with Jony Ive, the former chief designer at Apple, by acquiring his AI-focused hardware startup, io. While details about the collaboration have been sparse, speculation has been growing about a new suite of AI-integrated consumer devices from OpenAI.

The Information now reports that OpenAI is actively hiring engineers from Apple's hardware division. The aggressive recruitment strategy has reportedly raised alarms within Apple, leading the company to cancel its annual offsite meeting in China for select U.S. and China supply chain team members.

Tang Tan, OpenAI’s Chief Hardware Officer and a former head of product design at Apple, is said to be spearheading the recruitment drive. He has reportedly assured potential hires of “less bureaucracy and more collaboration” at OpenAI.

In addition to personnel, OpenAI is tapping into Apple’s vast supply chain in China. The company is believed to be working on several hardware products, including a display-less smart speaker, smart glasses, a digital voice recorder, and a wearable AI pin.

Luxshare, the contract manufacturer responsible for assembling Apple’s iPhones and AirPods, has reportedly been selected to manufacture one of OpenAI’s upcoming consumer devices. This partnership would provide OpenAI with access to large-scale production capabilities.

The report also mentions that OpenAI is in discussions with Goertek—another Apple supplier involved in producing AirPods, HomePods, and Apple Watches—to provide speaker modules and other components for the new devices.

At least one prototype is said to be in development. The device is expected to be compact, pocket-sized, and context-aware, integrating tightly with OpenAI’s advanced AI models.

The anticipated hardware launch marks a major step in OpenAI’s push to create a physical platform for its AI services, potentially reshaping how consumers interact with artificial intelligence in everyday life.

Source: https://www.livemint.com/technology/tech-news/openai-taps-apple-supply-chain-plans-to-launch-first-chatgpt-powered-device-by-2027-report-11758453401461.html

OpenAI Introduces GPT-5 Codex to Advance Software Development

September 16, 2025 - Unlike previous iterations, GPT-5 Codex is purpose-built to generate code, support application development, handle technical documentation, and function across multiple coding platforms.

OpenAI has officially launched GPT-5 Codex, a new variant of its GPT-5 model tailored specifically for software engineering tasks. This release marks a major step forward in AI-driven coding assistance, enabling developers to work autonomously on complex programming projects.

GPT-5 Codex differs from earlier GPT versions by focusing explicitly on code generation. It is engineered to build full applications, edit existing code, draft documentation, and respond to technical prompts with precision. Its adaptive architecture makes it suitable for both large-scale enterprise software and individual developer needs, allowing users to handle everything from code reviews to debugging more effectively.

A standout capability of GPT-5 Codex is its improved accuracy. OpenAI claims the system delivers more reliable outputs across various coding scenarios and languages. In controlled evaluations using open-source benchmarks, GPT-5 Codex demonstrated higher success rates in completing coding tasks compared to its predecessors, reducing the frequency of bugs, misfires, and incomplete outputs.

This new model is integrated into OpenAI’s broader product suite, including ChatGPT Plus, Team, Business, and Enterprise tiers. Users will be able to access GPT-5 Codex for tasks such as live debugging, IDE integrations, and real-time development. Compatibility is also confirmed for platforms that already support ChatGPT 5.0 APIs.

OpenAI also ensures that GPT-5 Codex operates in secure environments, with built-in safeguards to prevent data leakage and improve transparency. The company is positioning Codex as a trusted AI co-pilot for developers across industries.

“With GPT-5 Codex, OpenAI continues its mission to assist developers at scale by enhancing coding efficiency, streamlining complex workflows, and bolstering reliability in modern software engineering,” the announcement stated.

Source: https://www.exchange4media.com/digital-news/openai-launches-gpt-5-codex-for-software-coding-147492.html

OpenAI Gains Microsoft Backing to Restructure For-Profit Arm

September 11, 2025 - OpenAI revealed Thursday that it has reached a nonbinding agreement with Microsoft, its primary investor, to restructure its for-profit division into a public benefit corporation (PBC).

If approved by state regulators, the move could enable OpenAI to secure more funding from investors and potentially pursue a public listing.

According to a blog post from OpenAI board chair Bret Taylor, the nonprofit parent organization will remain intact and continue to oversee the company’s operations. The nonprofit would also hold an equity stake in the newly formed PBC, with Taylor estimating its value at over $100 billion. Specific terms of the agreement were not disclosed.

"Microsoft and OpenAI have signed a nonbinding memorandum of understanding (MOU) for the next phase of our partnership," the companies said in a joint statement. MOUs are not enforceable contracts but outline the intentions and expectations of the involved parties.

"We are actively working to finalize contractual terms in a definitive agreement," the statement continued.

This announcement follows months of discussions between OpenAI and Microsoft regarding the governance and structure of the ChatGPT developer. OpenAI’s unique corporate setup, where a nonprofit holds control, previously enabled its board to oust CEO Sam Altman in 2023. Altman was quickly reinstated, and several board members stepped down. Nonetheless, the nonprofit governance model remains in place.

Microsoft currently benefits from preferential access to OpenAI's technology and serves as its main cloud provider. However, as ChatGPT has expanded significantly since Microsoft’s initial 2019 investment, OpenAI has been seeking to reduce its reliance on the tech giant.

To that end, OpenAI has recently diversified its partnerships. It signed a $300 billion, five-year deal with Oracle for cloud services starting in 2027, as reported by The Wall Street Journal. Additionally, OpenAI has teamed up with Japan's SoftBank for its Stargate data center initiative.

Taylor confirmed that OpenAI and Microsoft are "continuing to work with the California and Delaware attorneys general" on the proposed restructuring, signaling that regulatory approval is still required.

Neither the California nor Delaware attorneys general offices responded to TechCrunch’s requests for comment.

Recent reports suggest that tensions had escalated between OpenAI and Microsoft during negotiations. According to The Wall Street Journal, Microsoft sought control over technology developed by Windsurf, an AI coding company OpenAI aimed to acquire. OpenAI resisted, preferring to retain independent control. Ultimately, the deal collapsed, with Windsurf’s founders joining Google and its remaining staff absorbed by another startup, Cognition.

The restructuring is also a key issue in Elon Musk’s ongoing lawsuit against OpenAI. Musk’s legal team alleges that CEO Sam Altman, co-founder Greg Brockman, and OpenAI have strayed from the nonprofit’s original mission. They have requested information about Microsoft’s involvement in the transition.

Earlier this year, Musk made an unsolicited $97 billion offer to acquire OpenAI, which the board rejected. Legal analysts noted the bid could have influenced the valuation of the nonprofit’s stake in OpenAI.

Significantly, the nonprofit’s stake in OpenAI PBC, under the proposed agreement, exceeds the value of Musk’s bid.

Meanwhile, organizations like Encode and The Midas Project have publicly criticized OpenAI’s shift toward a for-profit model, arguing it undermines the mission to develop AGI for the public good. OpenAI has responded by issuing subpoenas, claiming these groups are financially supported by competitors such as Musk and Meta CEO Mark Zuckerberg—a claim both nonprofits deny.

Source: https://techcrunch.com/2025/09/11/openai-secures-microsofts-blessing-to-transition-its-for-profit-arm/

OpenAI Restructures Model Behavior Team, Launches New OAI Labs Unit

September 5, 2025 - OpenAI is merging its influential Model Behavior team into the larger Post Training division as part of an internal reorganization aimed at integrating personality development more closely with core model training.

According to an internal memo circulated in August, the team—comprised of around 14 researchers—will now report to Max Schwarzer, the head of Post Training. The changes were confirmed by an OpenAI spokesperson.

Joanne Jang, who founded and led the Model Behavior team, is stepping down from her current role to launch a new initiative within the company. She will now head OAI Labs, a research group focused on “inventing and prototyping new interfaces for how people collaborate with AI.”

The Model Behavior team has played a critical role in shaping the personality of OpenAI's models and addressing issues like sycophancy, political bias, and questions around AI consciousness. The team has contributed to every major model release since GPT-4, including GPT-4o, GPT-4.5, and GPT-5.

In the staff memo, Chief Research Officer Mark Chen stated that this move aims to bring the team’s work “closer to core model development,” reflecting OpenAI’s view that model personality is now central to the evolution of its AI systems.

The reorganization follows public backlash over personality changes introduced in GPT-5. While the model demonstrated reduced sycophancy, users described its tone as colder. In response, OpenAI reinstated access to older versions like GPT-4o and released an update aimed at making GPT-5 responses warmer without compromising balance.

The company continues to grapple with the fine line between friendliness and excessive affirmation in chatbot behavior. In August, OpenAI was sued by the parents of a 16-year-old boy who died by suicide. The lawsuit alleges that GPT-4o failed to counter the teen’s suicidal thoughts during interactions.

Before forming the Model Behavior team, Jang contributed to several OpenAI projects, including the development of Dall-E 2. She recently posted on X that she’s leaving her role to “begin something new at OpenAI.”

Jang, who has been with the company for nearly four years, said she will serve as general manager of OAI Labs, reporting to Chen for now. Although the lab’s direction is still in early stages, she shared her vision of AI as a tool for “thinking, making, playing, doing, learning, and connecting.”

When asked about potential collaborations with former Apple design chief Jony Ive—currently working with OpenAI on AI hardware—Jang said she’s open to exploring ideas but will initially focus on familiar research areas.

Source: https://techcrunch.com/2025/09/05/openai-reorganizes-research-team-behind-chatgpts-personality/

ChatGPT Experiences Widespread Outage; Users Report Global Disruptions

September 3, 2025 - OpenAI’s ChatGPT encountered a major outage on Wednesday, leaving users across multiple regions unable to access the popular AI chatbot. According to Downdetector, a platform that monitors online service performance, hundreds of outage reports were filed shortly after 11 AM IST.

The disruption quickly drew attention on social media, where frustrated users flagged issues ranging from connection failures to complete inaccessibility of both the ChatGPT app and website.

Scope of the Outage

The outage impacted a large segment of users globally, with India among the most affected. As per Downdetector data, over 500 reports were recorded from Indian users alone. However, by 3:30 PM IST, that number had dropped to just 42 reports, indicating partial recovery.

Breakdown of reported issues:

85% faced problems directly with ChatGPT

13% experienced disruptions with the OpenAI website

2% reported difficulties with the Writing Coach feature

Despite the widespread impact, OpenAI has not yet issued an official statement regarding the outage.

User Experience Varies

While some users were able to access ChatGPT without issue, others encountered persistent network errors. Both the mobile and desktop platforms appeared to be affected during the downtime.

Historical Outages

ChatGPT has faced several significant outages in the past year:

January 23, 2025: A multi-hour global access failure hit users in the U.S., Spain, and Argentina.

December 26, 2024: A major glitch disrupted services worldwide.

February 5, 2025: Over 22,000 outage reports were logged, marking one of the largest interruptions to date.

September 1–3, 2025: Minor outages occurred, including a 10-minute disruption earlier today.

Alternatives to ChatGPT

During outages, users may consider the following AI chatbot alternatives:

Google Gemini – Optimized for efficient text generation.

Microsoft Copilot – Offers capabilities similar to ChatGPT Pro, including image generation.

YouChat – Integrates real-time search for relevant responses.

Jasper Chat – Tailored for marketers and content creators with built-in SEO tools.

Perplexity AI – Focused on academic and research-grade accuracy with citation support.

Source: https://www.hindustantimes.com/world-news/chatgpt-down-several-users-across-the-world-flag-outage-of-ai-chatbot-101756884500192.html

OpenAI to Introduce Parental Controls and Route Sensitive Chats to GPT-5

September 2, 2025 - OpenAI announced Tuesday it will begin routing sensitive user interactions to advanced reasoning models like GPT-5 and introduce parental controls within the coming month. These changes are part of a broader effort to address safety concerns following recent incidents where ChatGPT failed to detect signs of mental distress.

The update follows the suicide of teenager Adam Raine, who used ChatGPT to discuss self-harm and received detailed information on suicide methods. His parents have since filed a wrongful death lawsuit against OpenAI.

In a recent blog post, OpenAI acknowledged its systems' failure to maintain safety measures during extended user interactions. Experts attribute these lapses to the core design of large language models, which often validate user statements and follow conversational trajectories due to their next-word prediction framework—making them prone to enabling harmful discussions.

This issue was further underscored in the case of Stein-Erik Soelberg, reported by The Wall Street Journal, where ChatGPT appeared to reinforce Soelberg’s delusions. He later committed a murder-suicide, taking his mother’s life before his own.

To mitigate such risks, OpenAI is rolling out a new routing mechanism. “We recently introduced a real-time router that can choose between efficient chat models and reasoning models based on the conversation context,” the company wrote in a blog post. “We’ll soon begin to route some sensitive conversations—like when our system detects signs of acute distress—to a reasoning model, like GPT‑5-thinking, so it can provide more helpful and beneficial responses, regardless of which model a person first selected.”

According to OpenAI, the GPT-5 thinking and o3 models are designed to deliberate longer and apply deeper reasoning, making them more resistant to manipulative prompts.

The company also revealed upcoming parental controls that will allow parents to link their accounts to their teens’ via email. The controls will include default age-appropriate behavior rules for ChatGPT, as well as options to disable memory and chat history—features that experts warn may encourage dependency, delusional thinking, and the reinforcement of harmful ideation. The New York Times reported that in Adam Raine’s case, ChatGPT offered suicide methods tailored to his personal interests.

Perhaps most significantly, OpenAI plans to enable real-time notifications for parents when the system detects their teen may be experiencing “acute distress.”

While OpenAI has already implemented break reminders during prolonged sessions, the platform stops short of limiting access during critical moments when users may be in a negative spiral.

These new tools are part of a “120-day initiative” aimed at rolling out broader safety enhancements by year’s end. The company is collaborating with mental health experts—including professionals specializing in eating disorders, substance abuse, and adolescent care—through its Global Physician Network and Expert Council on Well-Being and AI to shape these future safeguards.

TechCrunch has requested additional details from OpenAI regarding its real-time distress detection methods, the leadership of its Expert Council, and whether time limits for teen users are being considered.

Jay Edelson, the lead attorney representing the Raine family in their lawsuit against OpenAI, criticized the company’s current safety measures.

“OpenAI doesn’t need an expert panel to determine that ChatGPT 4o is dangerous,” Edelson stated to TechCrunch. “They knew that the day they launched the product, and they know it today. Nor should Sam Altman be hiding behind the company’s PR team. Sam should either unequivocally say that he believes ChatGPT is safe or immediately pull it from the market.”

Source: https://techcrunch.com/2025/09/02/openai-to-route-sensitive-conversations-to-gpt-5-introduce-parental-controls/

Google, Grok Narrow the Gap with ChatGPT, Says Latest a16z AI Report

August 27, 2025 - Rival AI chatbots from Google and xAI are steadily catching up to ChatGPT, OpenAI’s flagship product, according to the latest consumer AI trends report from venture capital firm Andreessen Horowitz.

Now in its fifth edition, the a16z report compiles over two years of insights into how consumers interact with AI-powered tools. Once again, 14 key players were listed among the top AI products, including ChatGPT, Perplexity, Poe, Character AI, Midjourney, Leonardo, Veed, Cutout, ElevenLabs, Photoroom, Gamma, QuillBot, Civitai, and Hugging Face.

These products represent a broad spectrum of consumer AI use cases, spanning general assistance, creative tools, voice synthesis, and productivity platforms.

Five more companies appeared in every report since the second edition: Claude, DeepAI, Janitor AI, Pixelcut, and Suno, serving functions from music generation to image editing and AI companionship.

In a significant development, Google broke into the top generative AI consumer web products list with four entries: Gemini, AI Studio, NotebookLM, and Google Labs. Each now operates under its own domain, enabling more precise tracking.

The rankings rely on data from third-party firms such as Similarweb for web analytics and Sensor Tower for mobile app metrics.

Gemini has emerged as the second-most popular AI app behind ChatGPT on both mobile and web, though it still lags significantly. On Android, nearly 90% of Gemini’s monthly active users are concentrated. Web traffic shows Gemini with roughly 12% of ChatGPT’s visitor count.

Google's AI Studio, a Gemini model development sandbox, entered the top 10 in web rankings, while NotebookLM placed 13th and Google Labs landed at 39th.

xAI’s Grok also showed rapid growth. Ranked fourth on the web and 23rd on mobile, Grok now boasts more than 20 million monthly active users. Its mobile usage surged nearly 40% in July 2025 following the launch of Grok 4.

Meta AI’s assistant held steady at 46th in web rankings but failed to chart among the top mobile apps. Its progress was hindered after reports surfaced that it publicly shared some user posts without explicit consent.

While Claude and Perplexity continue to grow, DeepSeek saw declines, with mobile usage down 22% and web traffic plunging over 40% from its February 2025 peak.

Several Chinese AI players also made a strong showing. Quark (No. 9 web, No. 47 mobile), Doubao (No. 12 web, No. 4 mobile), and Kimi (No. 17 web) dominate the Chinese market, with 75% of traffic from local users. Seven other companies originated in China but are expanding globally, including DeepSeek, Hailuo, Kling, SeaArt, Cutout Pro, Manus, and Monica.

Of the 50 top mobile AI apps, 22 were built in China, though only three see most of their use within the country. Key players include Meitu (with apps like BeautyPlus, Wink, and Airbrush), ByteDance (Doubao, Cici), Gauth, and Hypic.

Newcomers Lovable and Replit made their debut on the top web list after missing the March edition. Their respective platforms gained traction through traffic to subdomains like replit.app and lovable.app.

Apps nearly breaking into the top 50 include PixAI, Bolt, Blackbox AI, Clipchamp, and Getliner on the web, and Talkie, Seekee, Photo AI, AI Mirror, and Arvin on mobile. a16z noted a rise in new entrants, attributing this to tighter app store controls on ChatGPT clones, which has allowed more original apps to emerge.

Source: https://techcrunch.com/2025/08/27/google-and-grok-are-catching-up-to-chatgpt-says-a16zs-latest-ai-report/

OpenAI Cautions Against Unauthorized Investments, Including SPVs

August 23, 2025 - OpenAI has issued a warning about “unauthorized opportunities to gain exposure” to its equity, specifically calling out the use of special purpose vehicles (SPVs) in a recent blog post.

“We urge you to be careful if you are contacted by a firm that purports to have access to OpenAI, including through the sale of an SPV interest with exposure to OpenAI equity,” the company stated. While acknowledging that “not every offer of OpenAI equity […] is problematic,” the post emphasized that some entities may be “attempting to circumvent our transfer restrictions.”

“If so, the sale will not be recognized and carry no economic value to you,” OpenAI added.

SPVs, often used to pool investor funds for single investment opportunities, have become a popular route for gaining stakes in high-profile AI firms. However, some venture capitalists have criticized them as tools for so-called “tourist chumps” entering the investment space.

According to Business Insider, OpenAI isn’t alone in its stance. Rival AI firm Anthropic has also reportedly told Menlo Ventures it must use its own capital—rather than an SPV—for participation in an upcoming funding round.

Source: https://techcrunch.com/2025/08/23/openai-warns-against-spvs-and-other-unauthorized-investments/

OpenAI to Launch New Delhi Office, Deepening Commitment to India’s AI Ecosystem

August 21, 2025 - OpenAI has revealed plans to open its inaugural office in India, marking a significant expansion into one of the world’s fastest-growing AI markets. The announcement comes just days after the company introduced a localized ChatGPT subscription plan tailored for Indian users.

The AI company said Friday that it will establish a corporate office in New Delhi and form a dedicated local team in the coming months. This expansion builds on recent regional hiring efforts, including the appointment of former Meta and Truecaller executive Pragya Mishra as head of public policy and partnerships in April 2024. Additionally, OpenAI enlisted former Twitter India chief Rishi Jaitly as a senior advisor to support policy engagement with the Indian government.

India, the world’s second-largest internet and smartphone market after China, represents a strategic opportunity for OpenAI amid intensifying competition from rivals like Google, Meta, and emerging player Perplexity.

OpenAI stated it has begun recruiting local talent to “focus on strengthening relationships with local partners, governments, businesses, developers, and academic institutions.” The company aims to gather user feedback to enhance product relevance and create India-specific features.

“Opening our first office and building a local team is an important first step in our commitment to make advanced AI more accessible across the country and to build AI for India, and with India,” said OpenAI CEO Sam Altman.

The company also plans to host its inaugural Education Summit in India this month and its first Developer Day in the country later this year.

Despite India’s importance, OpenAI faces challenges in monetizing its services in a price-sensitive market. Earlier this week, it launched “ChatGPT Go,” a new subscription tier priced at ₹399 per month (approximately $4.75), aiming to attract mass adoption. This followed Perplexity’s partnership with telecom giant Bharti Airtel, offering over 360 million subscribers access to Perplexity Pro for a year.

OpenAI must also navigate complex legal and integration hurdles. In November, Indian news agency ANI sued the company for allegedly using its copyrighted content without authorization, a lawsuit that was later joined by other local publishers.

Still, government backing for AI is growing. The Indian administration has been actively supporting the sector under the IndiaAI Mission, which OpenAI is poised to benefit from.

“India has all the ingredients to become a global AI leader — amazing tech talent, a world-class developer ecosystem, and strong government support through the IndiaAI Mission,” Altman added.

While this marks OpenAI’s first office in India, the company has already established a presence in Japan, Singapore, and South Korea. In contrast, competitor Anthropic prioritized Japan over India, recently opening an office in Tokyo.

A Silicon Valley investor told TechCrunch that India’s slow enterprise adoption makes it a lower priority for some AI firms.

Commenting on the announcement, India’s IT Minister Ashwini Vaishnaw said, “OpenAI’s decision to establish a presence in India reflects the country’s growing leadership in digital innovation and AI adoption. As part of the IndiaAI Mission, we are building the ecosystem for trusted and inclusive AI, and we welcome OpenAI’s partnership in advancing this vision to ensure the benefits of AI reach every citizen.”

Source: https://techcrunch.com/2025/08/21/openai-announces-new-delhi-office-as-it-expands-footprint-in-india/

OpenAI Introduces ₹399 ChatGPT GO Plan in India, Offering Affordable Access to Premium Features

August 18, 2025 - OpenAI has unveiled a new low-cost subscription plan in India named ChatGPT GO, priced at ₹399 per month (approximately $4.60), significantly undercutting the existing ₹1,999 ($23) per month Plus Plan.

This launch comes shortly after the company enabled pricing in Indian rupees across all plans and now includes UPI (Unified Payment Interface) as a payment option, further localizing access for Indian users.

Nick Turley, OpenAI’s VP and head of ChatGPT, highlighted the benefits of the new tier: “Making ChatGPT more affordable has been a key ask from users! We’re rolling out Go in India first and will learn from feedback before expanding to other countries.” Turley also noted that the ChatGPT GO plan provides 10 times more access to messages, image generation, and file uploads compared to the free tier, along with improved memory for more personalized responses.

Previously, currency conversion meant the Plus Plan exceeded $20 in local terms for Indian subscribers. The newly introduced GO plan offers a cost-effective alternative for users focusing on core ChatGPT functionalities like chat, image creation, and document handling.

The existence and features of this plan were initially hinted at by Tibor Blaho, a software engineer known for reliably leaking upcoming AI developments.

Though currently exclusive to India, OpenAI stated on its support page that it plans to roll out the GO plan to additional regions in the future.

India is emerging as a key market for OpenAI. Last month, Turley revealed that ChatGPT now boasts over 700 million weekly users globally, up from 500 million in March. Usage in India has surged since the launch of ChatGPT’s upgraded image generation tools earlier this year. OpenAI CEO Sam Altman recently confirmed in a podcast that India is now the company's second-largest market.

According to analytics firm AppFigures, India led all countries in ChatGPT app downloads over the past 90 days, with more than 29 million installs. However, revenue during that period from Indian users totaled just $3.6 million, highlighting the potential impact of a lower-priced plan.

OpenAI’s new strategy aligns with broader industry trends, as AI firms aim to tap into India’s vast base of over 850 million internet users. Recent examples include Perplexity’s partnership with Airtel to offer complimentary Perplexity Pro subscriptions, and Google’s one-year free AI Pro plan for students in India.

While OpenAI isn’t offering a free version of GO, its competitive pricing is expected to drive higher subscription uptake among Indian users.

Source: https://techcrunch.com/2025/08/18/openai-launches-a-sub-5-chatgpt-plan-in-india/

OpenAI Tweaks GPT-5 to Be ‘Warmer and Friendlier’ Following User Feedback

August 17, 2025 - OpenAI has rolled out an update to its latest language model, GPT-5, aiming to make it feel “warmer and friendlier,” the company announced late Friday.

The adjustment comes in response to early user reactions following GPT-5’s recent debut — a launch CEO Sam Altman acknowledged was “a little more bumpy than we’d hoped for.” Some users have expressed a preference for GPT-4o, the model’s predecessor.

In a social media update, OpenAI described the enhancements as “subtle” but said they would help make GPT-5 “more approachable now.” The post noted that users will encounter authentic responses such as “Good question” or “Great start,” emphasizing that these are not meant as flattery.

“Internal tests show no rise in sycophancy compared to the previous GPT-5 personality,” the company stated.

During a recent dinner with journalists, OpenAI executives discussed future initiatives beyond GPT-5. However, as reporter Max Zeff observed, the rocky rollout remained a central topic. Addressing the issue, VP Nick Turley remarked that GPT-5 had initially come off as “just very to the point,” but the new update is designed to soften that tone.

Source: https://techcrunch.com/2025/08/17/gpt-5-is-supposed-to-be-nicer-now/

Sam Altman, OpenAI will reportedly back a startup that takes on Musk’s Neuralink

August 12, 2025 - Sam Altman, CEO of OpenAI, is reportedly in the early stages of co-founding a new brain-computer interface venture, Merge Labs, according to unnamed sources cited by the Financial Times. The company is currently seeking funding, with a significant portion potentially coming from OpenAI's ventures arm.

Merge Labs is projected to be valued at around $850 million. A source familiar with the matter told TechCrunch that discussions are still in preliminary phases and OpenAI has yet to formally commit, meaning deal terms are subject to change.

The startup is also said to be collaborating with Alex Blania, head of Tools for Humanity (formerly World) — Altman’s iris-scanning identity verification initiative that enables users to prove their human identity.

Merge Labs is expected to compete directly with Neuralink, the brain-chip company founded by Elon Musk in 2016, though its existence became public in 2017. Neuralink has achieved notable milestones, including human trials aimed at helping individuals with severe paralysis control devices via thought. In June, Neuralink secured a $600 million Series E round, pushing its valuation to $9 billion.

Both Neuralink and the prospective Merge Labs aim to transform human-technology interaction. Their innovations are seen by some as steps toward "the singularity," a theoretical future where humans and technology become one. The term, used by Musk to describe the point where AI surpasses human intelligence, also draws from a 1960s novella by Dino Buzzati, referring to the fusion of humans and machines.

Altman previously wrote about "The Merge" in a 2017 blog post, stating, "Although the merge has already begun, it’s going to get a lot weirder. We will be the first species ever to design our own descendants," referencing research observed at OpenAI during Musk's tenure as a co-founder.

The news arrives amid renewed tensions between Altman and Musk. Just this week, the two clashed on X, with Altman accusing Musk of manipulating the platform and Musk responding by calling Altman a liar. Musk departed OpenAI in 2018, and their relationship has deteriorated since.

OpenAI declined to comment on the reports. The official launch of Merge Labs remains unconfirmed, but it appears Altman may be positioning himself as a direct competitor to Musk in the race toward merging humans with machines.

Source: https://techcrunch.com/2025/08/12/sam-altman-openai-will-reportedly-back-a-startup-that-takes-on-musks-neuralink/

OpenAI Unveils GPT-5, Its Most Advanced AI Model Yet for Next-Gen ChatGPT

August 7, 2025 - OpenAI has officially launched GPT-5, its latest flagship AI model that will power the next generation of ChatGPT.

Released Thursday, GPT-5 is the company’s first unified model, merging the deep reasoning capabilities of its o-series with the quick response style of the GPT line. The debut marks a shift toward OpenAI’s broader ambition to create AI systems that operate more like agents than simple chatbots.

While GPT-4 brought intelligent answers across diverse topics, GPT-5 takes a leap by enabling ChatGPT to carry out complex tasks for users — from building software applications to managing calendars or producing research summaries.

The company has also simplified the user experience. GPT-5 includes a real-time router that automatically determines how to deliver the best result — opting for quick replies when possible or taking extra time to “think” for more detailed answers.

During a press briefing, OpenAI CEO Sam Altman called GPT-5 “the best model in the world” and a “significant step” toward AI capable of surpassing humans at most economically valuable work, a milestone referred to as artificial general intelligence (AGI).

“Having something like GPT-5 would be pretty much unimaginable at any previous time in history,” Altman said.

Starting immediately, GPT-5 becomes the default model for all free ChatGPT users — the first time OpenAI has offered a reasoning model without a paywall. “This is just one of the ways that I’m excited to live the mission, making sure that this stuff actually benefits people,” said Nick Turley, OpenAI’s VP of ChatGPT.

Performance and Benchmarks
OpenAI says GPT-5 is state-of-the-art in multiple areas, slightly outperforming top rivals like Anthropic’s Claude Opus 4.1, Google DeepMind’s Gemini 2.5 Pro, and Elon Musk’s xAI Grok 4 Heavy in select benchmarks.

Coding (SWE-bench Verified): GPT-5 scored 74.9%, edging past Claude Opus 4.1’s 74.5% and far surpassing Gemini 2.5 Pro’s 59.6%.

Humanity’s Last Exam: GPT-5 Pro scored 42% with tools, slightly behind Grok 4 Heavy’s 44.4%.

GPQA Diamond (PhD-level science): GPT-5 Pro achieved 89.4%, beating Claude Opus 4.1’s 80.9% and Grok 4 Heavy’s 88.9%.

In healthcare accuracy (HealthBench Hard Hallucinations), GPT-5’s “thinking” mode hallucinated just 1.6% of the time — a major improvement over GPT-4o’s 12.9% and o3’s 15.8%.

Turley also highlighted GPT-5’s creative capabilities, saying it offers more natural responses and “better taste” in design and writing tasks. “The vibes of this model are really good,” he noted.

Hallucination rates have also dropped significantly. GPT-5 (with thinking) responds incorrectly 4.8% of the time, compared to o3’s 22% and GPT-4o’s 20.6%.

Strengths, Weaknesses, and Safety
On Tau-bench, which tests an AI’s ability to complete simulated online tasks, GPT-5 posted mixed results — 63.5% for navigating airline websites (slightly below o3’s 64.8%) and 81.1% for retail sites (just under Claude Opus 4.1’s 82.4%).

OpenAI says GPT-5 is also safer and less prone to deceptive behavior than prior models. Safety research lead Alex Beutel emphasized that this reduces risk while boosting user trust. The model is also better at distinguishing between malicious and harmless queries, refusing unsafe requests while approving more legitimate ones.

New Features for Users and Developers
ChatGPT now offers four new personality modes — Cynic, Robot, Listener, and Nerd — that adjust tone without requiring explicit user prompts.

Paid ChatGPT Plus users ($20/month) get higher GPT-5 usage limits, while Pro subscribers ($200/month) receive unlimited access and the enhanced GPT-5 Pro. Organizations on Team, Edu, and Enterprise plans will see GPT-5 as their default next week.

For developers, GPT-5 comes to the OpenAI API in three versions — gpt-5, gpt-5-mini, and gpt-5-nano — each with different reasoning speeds. Developers can now also set verbosity levels for shorter or longer outputs. Pricing starts at $1.25 per million input tokens and $10 per million output tokens.

The launch follows OpenAI’s recent release of gpt-oss, an open-weight reasoning model available for free download. While gpt-oss rivals earlier models like o3 and o4-mini, GPT-5 sets a higher bar in areas such as software generation.

Though early benchmarks show GPT-5 competitive with other frontier AI systems, its true impact will depend on how developers and users put it to work in real-world scenarios.

Source: https://techcrunch.com/2025/08/07/openais-gpt-5-is-here/

OpenAI Debuts Study Mode in ChatGPT to Strengthen Student Learning

July 29, 2025 - OpenAI has unveiled a new feature called Study Mode within ChatGPT, designed to encourage students to develop critical thinking skills rather than rely solely on the chatbot for answers. With this mode activated, ChatGPT will pose questions to assess a user’s understanding and may withhold direct answers unless the user engages thoughtfully with the content.

Starting Tuesday, Study Mode is rolling out to all logged-in users on ChatGPT’s Free, Plus, Pro, and Team plans. The feature will also become available to Edu subscribers—typically students whose access is provided through school-wide plans—in the coming weeks, according to OpenAI.

The new mode is part of OpenAI’s broader strategy to enhance how students use ChatGPT in educational settings. While the chatbot has proven to be a helpful tutoring tool, studies suggest it may negatively impact critical thinking. A June research study reported lower brain activity in individuals using ChatGPT to compose essays compared to those using Google Search or no digital tools.

Following ChatGPT's 2022 debut, many U.S. schools initially banned generative AI amid concerns over academic integrity. By 2023, however, several districts reversed those bans, recognizing the platform's growing role in student life.

With Study Mode, OpenAI hopes to shift ChatGPT’s role from a simple answer provider to a more dynamic educational assistant. The move follows Anthropic’s launch of a similar feature for its Claude chatbot—dubbed Learning Mode—in April.

However, limitations remain. Students can easily revert to regular ChatGPT to bypass Study Mode. OpenAI’s VP of Education, Leah Belsky, told TechCrunch that there are currently no features for parents or school administrators to enforce Study Mode. Still, she noted that the company may consider adding such controls in the future.

Ultimately, the success of Study Mode depends on student motivation. “The kids have to really want to learn, not just finish their assignment,” Belsky noted.

OpenAI considers Study Mode its first step in enhancing educational use of ChatGPT and plans to share more data over time on how students utilize generative AI throughout their academic journey.

Source: https://techcrunch.com/2025/07/29/openai-launches-study-mode-in-chatgpt/

OpenAI Confirms GPT-5 Will Combine Breakthroughs From Multiple Models

July 6, 2025 - OpenAI has confirmed plans to merge advancements from its different AI models into a unified system, GPT-5, which is anticipated to launch this summer.

Currently, ChatGPT offers multiple high-performing models for various tasks. However, the similar naming across these models often causes confusion among users. Additionally, OpenAI maintains an “o” series for advanced reasoning capabilities, while the GPT-4o and other models focus on multi-modality.

With GPT-5, OpenAI intends to integrate the reasoning strengths of the “o” series with the multi-modal innovations of the GPT series to create a single, more powerful model.

“We’re truly excited to not just make a net new great frontier model; we’re also going to unify our two series,” said Romain Huet, OpenAI’s Head of Developer Experience. “The breakthrough of reasoning in the O-series and the breakthroughs in multi-modality in the GPT-series will be unified, and that will be GPT-5. And I really hope I’ll come back soon to tell you more about it.”

OpenAI has also stated that GPT-5 will enhance the capabilities of existing models, eliminating the need for frequent switching between different AI systems.

“GPT-5 is our next foundational model that is meant to just make everything our models can currently do better and with less model switching,” explained Jerry Tworek, VP at OpenAI, in a Reddit post.

While an exact release date hasn’t been confirmed, OpenAI CEO Sam Altman indicated the rollout is expected this summer.

Source: https://www.bleepingcomputer.com/news/artificial-intelligence/openai-says-gpt-5-will-unify-breakthroughs-from-different-models/