[ctx]ctxseccontext security: attacks on the harness and context layer of AI agents, plus exploited-now news, hunting and playbooks.

← Industry News · markdown

confirmed (OPAQUE announcement)

Confidential-AI startup OPAQUE Systems winds down as its team joins Anthropic; open projects including Confidential MCP stay open

OPAQUE Systems announced on Oct 13 that its team is joining Anthropic and that the company is winding down. OPAQUE grew out of UC Berkeley's RISELab and built confidential-computing tech that lets AI workloads run on sensitive data under hardware-enforced protection. It later moved into 'verifiable AI', meaning cryptographic evidence of what ran, where it ran and under what policies. SecurityWeek describes the deal as an acquisition with undisclosed terms; OPAQUE's own post says only that the team is joining Anthropic. OPAQUE's site says its open-source work continues under AgenTrust: the Weight Custody Manifest, Agent Manifest, Confidential MCP, Confidential A2A and TRACE stay available under their existing licenses. The Confidential Computing Summit moves to the Linux Foundation.

Sources: OPAQUE: Next Chapter (Oct 13) · opaque.co · SecurityWeek