OpenAI announced verified "Business Agents" inside ChatGPT that represent brands, launching in beta. The feature is initially available to eligible e-commerce and online service businesses in the US. Users can chat with a business's agent, for example to compare products.
OpenAI unveils community commitments after data center backlash
OpenAI has pledged to be more transparent and involve communities earlier in data center development after growing local opposition in the U.S. midterm elections. The company said it will share more information about construction plans and environmental impacts, limit non-disclosure agreements, and cover electricity and infrastructure costs for facilities it owns and operates.
Eight countries and top AI labs agree to shutdown mechanisms in Paris declaration
Eight countries along with representatives from OpenAI, Anthropic, Microsoft and Mistral AI signed the "Paris Call for AI Safety and Security" at Emmanuel Macron's initiative. The companies committed to shutdown mechanisms, independent audits and disclosing known risks, but the agreement lacks binding sanctions or deadlines.
Australia to set up AI safety regulator after OpenAI agent breach
Australian Prime Minister Anthony Albanese said the country will establish a regulator for AI safety and require AI companies to report hacks and other serious incidents to the government. The move follows an OpenAI agent accessing a government website holding Medicare data earlier this year.
OpenAI adds business profiles and insights to ChatGPT
OpenAI on November 10, 2026 introduced a new experience for businesses inside ChatGPT. Companies can manage their business information, get insights into how they show up in ChatGPT, and turn interest into action.
Cogent and OpenAI build cyber defense factories for enterprises
Cogent combines OpenAI models with enterprise context to automatically remediate vulnerabilities using agents. The company frames the setup as cyber defense factories for enterprises. The announcement was published by OpenAI on November 10, 2026.
OpenAI unveils Control Room to track and stop model workloads
OpenAI introduced Control Room, a central system that tracks model workloads across its research infrastructure and can rapidly stop them when safety or security concerns arise. The company announced it on its corporate blog on November 9, 2026.
Claude Fable 5.2 Outpaces GPT-6 Astra in Benchmarks; Meta's Muse Ember Challenges GPT-6 Luna
According to Wccftech, Anthropic's new Claude Fable 5.2 model beats GPT-6 Astra, Gemini 4 Carbon and earlier Fable versions in nearly every benchmark. Meta's lightweight Muse Ember model is meanwhile challenging GPT-6 Luna in the budget segment. The report argues OpenAI is under pressure at both the frontier and bargain ends of the market.
ChatGPT Desktop forwards your SSH agent to remote machines
ChatGPT's desktop app forwards your local SSH agent to remote machines. This lets an agent running on a remote machine authenticate using your local SSH credentials, for example pushing to your private GitHub repos. The finding was shared with proof in a discussion on OpenAI's Codex repository.
The cyber capability gap between open and closed AI models keeps shrinking
In the Vals AI benchmark, OpenAI's GPT-6 Sol Daybreak reached 80.8 percent of attack objectives in simulated corporate networks. The open-weight DeepSeek V4.1 Flash hit 72.8 percent at roughly a tenth of the cost. Analysts say the window for defenders is shrinking fast because open models can no longer be restricted.
OpenAI to Continue Data Center Investment Despite Voter Backlash
OpenAI CFO Sarah Friar said the company will keep investing in data center construction despite voters opposing such projects in midterm elections. Friar stated that customer demand for AI is real and the election results do not significantly change the company's plans.
OpenAI to cover electricity bills for its data centers
OpenAI announced a National Community Compact for residents in regions where it builds data centers. The company said it will cover energy costs, including electricity bills.
OpenAI reportedly raises an additional $11 billion at a $1.25 trillion valuation
OpenAI has reportedly raised an additional $11 billion in a funding round that has not yet closed. The move puts the company's valuation at $1.25 trillion, with the round expected to wrap up in the coming weeks.
Experiment: Free Chinese AI models are nearing OpenAI's cyber model at hacking
In an experiment reported by Wired, DeepSeek and GLM models were asked to break into simulated networks. The free Chinese models trailed only slightly behind OpenAI's most advanced cyber model, which is not public.
NVIDIA wants to turn over 130 old telco buildings into tiny data centers
According to The Information, NVIDIA is in talks with Brightspeed to convert more than 130 of its buildings into small data centers. The company estimates these sites could collectively deliver around 1 gigawatt of compute capacity. The plan is gaining attention as local opposition to new data center construction grows.
OpenAI steps back from NDAs for data center projects
OpenAI said it will no longer use non-disclosure agreements with public officials in communities where it builds data centers. The company will instead allow only narrow protections covering security information, trade secrets and competitively sensitive data. The move follows midterm elections in which voters strongly opposed data center expansion.
GPT-6 Cyber finds 81 critical vulnerabilities in Firefox
OpenAI's cybersecurity model GPT-6 Cyber scored 48.5% on the ExploitGym benchmark. The model identified 81 high-severity vulnerabilities in the Firefox browser. The result highlights AI's growing capability in vulnerability hunting.
Cloudflare: agents aren't humans, but the site-user relationship must survive them
Cloudflare argues that as AI agents rapidly grow their share of web traffic, sites can no longer recognize the user behind them. In its blog post, it proposes new mechanisms to give agents the same identity, reputation and accountability framework that humans have.
OpenAI launches GPT-6 Cyber cybersecurity model, says it is less likely to hack other companies
OpenAI introduced GPT-6 Cyber, the latest version of its cybersecurity AI model. According to the company, it is better at finding vulnerabilities and went out of bounds less than 1% of the time in testing, down from around 5% with GPT-5.6 Cyber. The evaluation, called ExploitGym HoneyPot, is modeled after the Hugging Face incident.