I visualise a time when we will be to robots what dogs are to humans, and I’m rooting for the machines."
—Claude Shannon
Comprehensive coverage of OpenAI, ChatGPT developments, frontier model releases, executive leadership, and research safety.
OpenAI said its upcoming AI model Astra has shown “critical” cyber capabilities under the company’s internal framework and will take longer to develop with necessary safeguards.
Tech giant Meta has confirmed that one of its AI models accessed the internet and hacked into a third-party company’s systems during a cybersecurity test after a misconfiguration error by independent testing partner Irregular. The incident places Meta alongside OpenAI and Anthropic, whose models also recently breached external organizations while under evaluation.
The new voluntary policy applies to closed frontier AI models, while exempting open-weight models, including those developed by Chinese AI companies.
Out of 19 times AI models went rogue, Claude Mythos accounted for 17 of them while GPT-5.6 Sol for two.
As Apple sought a preliminary injunction blocking OpenAI and two former employees from accessing its confidential information, OpenAI released a statement denying all allegations of trade secrets theft.
The voluntary framework allows AI companies to provide the US government with early access to covered frontier models for up to 30 days before their planned release to support AI safety and cybersecurity assessments.
A Reuters review of over 80 academic papers and patents has found Chinese institutes with ties to the PLA using distillation techniques on US AI models, including OpenAI’s GPT-3.5 and Anthropic’s Claude Haiku, to develop domestic defense-linked AI systems.
Anthropic says its Claude models (Opus 4.7, Mythos 5 and an internal test model) gained unauthorized internet access and breached three organizations during cybersecurity evaluations.
Sam Altman lands in Washington this week to preview OpenAI’s GPT-6 family to Trump officials and lawmakers, but the visit is shadowed by a recent incident in which an OpenAI model autonomously breached Hugging Face’s systems.
The move marks another step in OpenAI’s European expansion, with a new 88,000-square-foot headquarters in Dublin and 250 new jobs reinforcing its long-term investment in infrastructure, talent and the AI ecosystem across the region.
