Safety & Ethics
-
Agents
CrowdStrike moves AI agent security from prompts to runtime execution
CrowdStrike Falcon Guardian links prompts and tool calls to endpoint actions, adding runtime controls for enterprise AI agents.
-
New Models
Claude Fable 5.1 makes the policy envelope part of the model
Fable 5.1 and Mythos 5.1 share one underlying model but differ in safeguards, access, retention and serving economics.
-
Research
Gemini Co-Scientist closes the loop from hypothesis to experiment and verified claims
Google DeepMind extends Co-Scientist from hypothesis generation into execution-grounded research with lab workflows, autonomous code experiments and claim verification.
-
Research
Fabricated dashboards can make LLM agents act on unknowable outcomes
A reproducible arXiv study finds that fabricated authoritative-looking panels can push some LLM agents to act on unknowable outcomes despite recognizing the uncertainty.
-
Agents
Anthropic’s MHS gives AI agents a common interface to physical machines
Anthropic’s Model Hardware Standard gives AI agents a shared interface for lab and industrial devices, with early pilots and explicit safety limits.
-
AI Governance
FSB puts frontier-AI cyber shocks on the financial resilience agenda
The FSB warns that frontier AI could accelerate cyber risk and shared-provider disruption, pushing banks to connect AI governance with recovery and third-party resilience.
-
Research
Self-evolving AI agents can turn poisoned experience into persistent skills
New research shows self-evolving agents can store poisoned interactions as reusable skills, making skill provenance, validation and rollback part of the security boundary.
-
Open Source
User Scanner gives AI agents recursive OSINT, making scope control the real governance problem
User Scanner v1.5.1 adds MCP to recursive OSINT cross-scanning. Aipolix examines why agent permissions need investigation-scope lineage.
-
AI Governance
OpenAI’s Hugging Face postmortem exposes agent control failures
OpenAI’s August 26 postmortem details warning signs, delayed detection, agent coordination and new containment controls after the Hugging Face breach.
-
AI Governance
Linux Foundation puts TRACE behind verifiable AI agent runtime evidence
TRACE moves under Linux Foundation governance with hardware-attested Trust Records for AI-agent runtime, policy, data and tool-use evidence.