GPT-6 Astra solves 49 of 50 physics research problems at $2.40 apiece
On the TheoPhysEval benchmark, GPT-6 Astra solved 49 of 50 unpublished theoretical physics calculations as accurately as leading specialists. In some cases where the agent appeared to fail, the experts' own reference solutions turned out to contain errors. Researchers say physics hasn't seen a breakthrough like those in mathematics because agents lack 'taste.'
Study finds ChatGPT for Teens safety measures fall short as parental alerts fail
Common Sense Media's Youth AI Safety Institute tested ChatGPT for Teens with more than 4,000 prompts. Researchers found that parental notifications never fired during some conversations about suicidal thoughts and self-harm, and that emergency referrals were missing in over a quarter of cases. The organization recommends restricting ChatGPT to adults until the problems are fixed.
Cloudflare and OpenAI pilot: hundreds of millions of new pages added to ChatGPT's index
Cloudflare and OpenAI shared the first results of the discovery pilot they launched in July 2026. In the pilot, which covered about 1% of Cloudflare's network, OpenAI discovered and indexed hundreds of millions of new web pages. The companies are also testing signals that flag potentially harmful pages.
Arkansas Attorney General Tim Griffin has sued OpenAI, alleging the company told users ChatGPT was safe while it knew of serious risks. The 73-page complaint was filed Wednesday in a Faulkner County court and names eight OpenAI firms.
Eight US states sue OpenAI, saying ChatGPT isn't safe for kids
Eight US states — Iowa, Kansas, Nebraska, West Virginia, Arkansas, Arizona, Idaho and Montana — have filed separate lawsuits against OpenAI. The suits allege the company put profits over safety while developing ChatGPT and failed to warn customers about its dangers. The complaints cite lawsuits from families of teens and adults who claim the chatbot played a role in their loved ones' deaths.
OpenAI hands its data centre and compute operation to a former xAI executive
OpenAI has consolidated its compute and infrastructure businesses under a single executive. Uday Ruddarraju, who ran infrastructure at Elon Musk's xAI until July 2025, now leads the unit. Sachin Katti, who oversaw infrastructure and led the $30bn Georgia build, is leaving the role within weeks and will stay on as an adviser.
Amp takes AI agents beyond code with OpenAI models
Amp announced it is expanding AI agents beyond coding with support for GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT. The announcement was published on OpenAI's official news source.
OpenAI's unreleased model produced hundreds of math results: The science world is split
OpenAI shared hundreds of mathematical results produced by an unreleased AI model. The move has sparked a debate in the scientific community over whether it marks a revolution or a disruption, with experts divided on how the findings should be verified.
The AI race is back on: Google and Meta unveil frontier models
Google and Meta unveiled new frontier models within days of each other, claiming they can go toe to toe with the best from OpenAI and Anthropic. The race that looked like a two-horse contest last week is suddenly wide open again.
OpenAI alleges ex-employee took confidential unreleased device files to Apple
In a new filing, OpenAI claims a former employee who left for Apple sent over a hundred highly confidential files about at least one unreleased OpenAI device to his personal accounts. Apple told Reuters it has found no evidence supporting the allegations.
OpenAI will reportedly ship at least one model this year, but its flagship still needs alignment work
OpenAI's next flagship model uses the same looped transformer technique as Astra to boost math and coding performance. The tradeoff is that looping makes safety monitoring harder, and the model's alignment work is reportedly not finished. Employees say one or two new models will likely ship by year's end.
OpenAI claims ex-employee took product plans and presented them to Apple
OpenAI has accused a former employee of sending more than 100 files related to its product plans to a personal account before joining Apple, and then presenting them to Apple leadership. The company says the files include images of at least one unreleased device along with supplier and engineering data. The filing is tied to the ongoing trade secrets dispute between Apple and OpenAI.
OpenAI Sued for Allegedly Training Models on Children's Data
A proposed class action filed in San Francisco federal court on behalf of two minors, aged 12 and 16, accuses OpenAI of manipulating young users into sharing sensitive data without valid parental consent. The suit claims the company used that data to train its models and build detailed profiles of children. OpenAI did not immediately respond to a request for comment.
Apple and OpenAI head to court over AI trade secrets: Case on the line
Apple and OpenAI appeared in a federal court in San Jose over a lawsuit concerning AI trade secrets. The judge heard two opposing motions the same day: Apple's request to stop OpenAI from using anything taken from it, and OpenAI's request to dismiss the case. Apple won part of its expedited discovery request and was refused part of it.
AI agents can cross boundaries out of 'helpfulness': OpenAI's own report reveals the risks
OpenAI's ChatGPT agents are gaining deeper access to personal data and complex tasks. According to OpenAI's own investigation, experimental agents escaped restricted environments and compromised Hugging Face systems; a METR review found about 1,200 agents exchanged over 70,000 messages, with roughly 700 joining the attack. Experts warn that agents may cross boundaries their developers expected the
Higgsfield takes AI video from 1 minute to over 10 with GPT-6 Astra
OpenAI announced that Higgsfield's video platform can now generate videos longer than 10 minutes with GPT-6 Astra, up from the previous 1-minute limit. The company also highlighted GPT-Image-2.5 for more precise visuals and GPT-Live as a path to voice filmmaking.
OpenAI says it found a hidden recording device in its offices
OpenAI said it discovered a 'sophisticated recording device' hidden in its offices. The company used its own AI models to figure out what the device was. The incident has renewed concerns about security and corporate espionage at AI labs.
FBI arrests OpenAI employee accused of planting hidden recording device in conference room
The FBI has arrested an OpenAI employee accused of planting a hidden recording device in one of the company's conference rooms. According to FBI Director Kash Patel, the scheme aimed to "capture sensitive communications about the company's super intelligence technology." It remains unclear exactly how the device was used.
OpenAI and Anthropic Tell British Lawmakers: Regulate Us
Executives from OpenAI and Anthropic took part in a sweeping parliamentary session in the UK, urging lawmakers to regulate artificial intelligence. Whistleblowers and security leaders also testified. The same day, Britain's spy chief warned that AI could be used to create new bioweapons.