Anthropic Says China’s GLM-5.3 Nearly Matches Mythos at Hacking
Anthropic's red team says Zhipu's open-weight model built a Chrome exploit chain for about US$20 of compute and can have its refusals…
Breaches, zero-days, state hackers and the privacy tradeoffs baked into new tech. We name the files, do the math on the fines, and explain what each incident means for the people whose data was in it.
Anthropic's red team says Zhipu's open-weight model built a Chrome exploit chain for about US$20 of compute and can have its refusals…
Transluce says AI agents probed Library and Archives Canada for 1905-era divorce records. Canada's public account is one sentence; Australia's is a…
AI agents now log in, click and buy for you. A 30-minute pre-flight checklist so a hijacked agent can't take your whole…
A Pentagon records unit exposed SSNs and military job details for about 3 million people, and didn't notice for ten months. The…
OpenAI and Anthropic are probing tens of thousands of agent security incidents. The definitions, the counting and the disclosure lag are the…
Sweden's IMY fined the HR software supplier SEK 1.8 million after a breach of 2.2 million people's records. The decision shows that…
A breach reported by 404 Media itemizes the datasets behind Suno's models, weeks before discovery closes in the major labels' Boston lawsuit.
CAA and SAG-AFTRA killed Instagram's AI remix feature in three days. The derivative-works license that made it possible is untouched.
The EU and UK say FSB Centre 16 attacked Poland's grid to cut winter power to some 500,000 people, and sanctioned the…
Four breaches in three days took keys, tokens, license numbers and passwords. Identity is the new target, and it's the layer nobody…