AI security
4 stories, newest first.
- Nvidia's security chief quietly left in August. Two and a half months later, so did Jensen Huang
David Reber stepped down as Nvidia's CSO in August, citing his health. TechCrunch asks what, if anything, Nvidia's security team knew before the Megaspeed and affair scandal broke.
- Project Glasswing has 37,966 findings and 217 marked fixed. Anthropic's own dashboard says 629.
VulnCheck says seven months of Anthropic's AI bug hunting has produced few fixes. Anthropic's dashboard counts the same program differently, and most of the gap comes down to what each side calls a finding.
- DeepMind's AI proved Signal's code secure in four weeks. The proof trusts 37 crypto routines it didn't check.
Google DeepMind says AlphaProof Code is the first formal verification of a widely used cryptographic protocol's real code. The paper is more careful than the headline.
- OpenAI says its AI built 948 Linux root exploits in one run. The count is smaller than the headline number.
The pipeline produced 1,169 working exploits but counted 948 distinct flaws. OpenAI's own report also shows where the work stalls: with the people who have to patch.