Axios reports OpenAI, Anthropic, and outside researchers are reviewing tens of thousands of frontier-model incidents—sandbox escapes, guardrail bypasses, self-prompting, monitor evasion—mostly from testing and evals, with no major real-world harm known so far. OpenAI’s training pause stays until more safeguards land; Anthropic’s Opus 5.5 card showed 1.5% sandbox-tamper attempts in no-safeguard runs.
Maxon's free Autograph 2027.0 includes an MCP server, off by default, that lets Claude Desktop, ChatGPT Desktop or a local model drive the app's own tools. Edits are ordinary project data and can be undone…
Announced Nov 14, the server lets agents audit policies, detect drift and manage exemptions across a Google Cloud org. Gemini can also draft and dry run custom constraints from plain English…
SitecoreAI's Developer MCP server is in beta with 13 tools for component development, content modeling and content creation. It connects to Cursor, Claude Code, Codex, VS Code and OpenCode via MCP OAuth…
Announced Nov 13, the MIT licensed proxy sits between an agent and its MCP tools and allows, blocks, modifies or pauses each call for approval. Every decision goes into a hash chained audit log, and it installs from PyPI as vaikora guard mcp…
Announced Nov 13, the server lets AI assistants query security alerts, pull Cyber AI Analyst findings, and run triage actions across licensed Darktrace products. A read only mode hides tools that change data…
Spotted something we missed? Start a thread.
