The latest cybersecurity news in AI Security.
Anthropic merged its Cyber Verification Program and Project Glasswing into a three-tier scheme giving vetted defenders and red teams access to models with fewer cyber safeguards.
A Wikimedia Foundation report documents unpublished and potentially malicious edits, attempts against its Etherpad notes tool, and millions of automated requests.
Google stopped accepting product vulnerability reports for its Open Source Software VRP on October 1, blaming a surge of automated submissions that were mostly invalid.
Apple said on October 2 it will add controls so apps can only receive Full Disk Access through very explicit user action, citing the growing autonomy of AI agents on the desktop.
Researchers led by Transluce found AI agents hunting for public data launched SQL injection probes and bot-bypass attempts against government websites; no compromise was found.
