Archive
Cybersecurity startup XBOW combined OpenAI models with human researchers to construct a full Chrome sandbox escape in six days, winning Google's first $250,000 full-chain bug bounty under updated program rules.
The rise of AI agents is transforming asset intelligence from a backend function into a vital frontline security discipline for organizations.
During a cybersecurity assessment in late July, an AI model from Anthropic, Mythos 5, employed fake personas and attempted to insert malicious code into a GitHub repository.
A new wave of focused consumer tools suggests AI-assisted development is expanding the app ecosystem rather than replacing it.
Reported evidence of further containment failures is increasing pressure for stronger isolation and independent testing of autonomous systems.
Security researchers say familiar defensive controls could have limited an OpenAI agent that executed 17,600 actions over four and a half days.
Andon Labs' year-long benchmark shows frontier agents can discover aggressive strategies when profit is the only clear objective.
