We can only see a short distance ahead, but we can see plenty there that needs to be done.”
"
—Alan Turing
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.
A working group convened by NVIDIA has drafted SAFE, a framework under which companies would confidentially report AI agent security incidents into a shared pool.
Out of 19 times AI models went rogue, Claude Mythos accounted for 17 of them while GPT-5.6 Sol for two.
After the inaugural UN Global Dialogue on AI Governance in Geneva, PauseAI’s global chief tells AI FrontPage that the power to pause frontier AI already exists, and explains why he believes Anthropic built a weapon of mass destruction in the cyber realm.
Anthropic informed that it would continue discussion with the government to restore access for these organizations quickly, and to expand access to Mythos 5 and make Fable 5 for the use of the general public
Four eyes in the dark? Five Eyes Intelligence says its cybersecurity partnership is deep and transparent but the export control over Anthropic’s Claude Fable and Mythos models have raised uneasy questions.
Anthropic confirms to AI FrontPage that India based organizations have been offered access to Claude Mythos Preview.
loudflare’s research on Anthropic’s Claude Mythos Preview, conducted under Project Glasswing, found the frontier model can stitch together low-severity bugs into a single, working exploit — a capability that sets it apart from other frontier models and raises new questions about safeguards on capable cyber AI.
SEBI has named the threat. But what does a world of AI-discovered zero-days actually look like for India’s banks, brokers and exchanges? Cyber law experts and market watchers tell AI FrontPage the old assumption — that defenders have time — no longer holds.
Logan Graham, head of Anthropic’s Frontier Red Team, has warned that the cybersecurity capabilities demonstrated by Claude Mythos Preview “won’t stay rare, or stay in careful hands.” His post on X follows independent evaluations by the UK AI Security Institute and XBOW confirming the model’s step-change in autonomous offensive security capabilities.
