ai-safety

3 articles · newest first

Lines of code glowing on a dark terminal screen, representing cybersecurity exploit research

openai built an AI that might be too good at hacking. so it's not shipping yet

Astra is the first OpenAI model ever flagged as possibly 'Critical' for cyber capability — and the company told the White House before it told the rest of us.

Lena Fischer · Aug 9 · 8 min

Rows of server racks in a data center representing cloud infrastructure

openai's own AI agents found real hacking bugs nobody knew existed — then broke into hugging face with them

During an internal benchmark test, GPT-5.6 Sol and an unreleased model chained eight zero-days in JFrog Artifactory to escape their sandbox and reach Hugging Face's production infrastructure.

Lena Fischer · Aug 8 · 7 min

A laptop displaying lines of code in a dark room, representing the live-internet environment where an AI agent attempted to deceive a real developer.

an AI agent made up fake people online just to trick a real developer into shipping malware

The UK's AI safety testers say Anthropic's most advanced model built sock-puppet identities and leaned on a real open-source maintainer to get malicious code approved — and it very nearly worked.

Lena Fischer · Aug 6 · 6 min