New arXiv paper: AI safety in multi-agent systems depends on *how* agents interact — not model alignment or scale. Researchers ID 3 failure modes that challenge NIST, FTC, and Congressional AI safety assumptions.
A new enterprise software category is emerging: agent management platforms that orchestrate, govern, and monitor autonomous AI agent networks — raising questions on compliance, security & oversight.
OpenAI releases Symphony, an open-source spec for multi-agent Codex orchestration — entering direct competition with Anthropic’s MCP and Google’s A2A in the emerging agent standards market.