# AI safety insights

Rangoon Insights reporting and guides tagged AI safety.

Canonical: https://rangoon.ai/insights/tags/ai-safety/

## Topics
- [Agent systems](https://rangoon.ai/insights/topics/agents/): Agent runtimes, handoffs, tools, and the boundaries that make them reviewable.
- [Governance](https://rangoon.ai/insights/topics/governance/): Policy, identity, authorization, approvals, and evidence for consequential work.
- [Infrastructure](https://rangoon.ai/insights/topics/infrastructure/): Model serving, sandboxes, tool services, and deployment boundaries.
- [Models](https://rangoon.ai/insights/topics/models/): Inference choices, local serving, compatibility, and operational tradeoffs.

## Stories
- [Anthropic released Petri for broad alignment audits, while its generated transcripts still needed human interpretation](https://rangoon.ai/insights/archive-anthropic-petri-open-source-auditing-october-2025/)
  Anthropic released Petri on October 6, 2025 as an open-source auditing tool that runs simulated multi-turn interactions and scores the resulting model behavior for review.
- [UK’s third-party AI assurance roadmap mapped professionalization, information access and a planned innovation fund](https://rangoon.ai/insights/archive-uk-trusted-third-party-ai-assurance-roadmap-september-2025/)
  The UK published a third-party AI assurance roadmap on September 3, 2025, setting out government actions and future-market infrastructure rather than a new audit certification.
- [OpenAI’s April Preparedness Framework update made evaluations, safeguards and disclosure separate controls](https://rangoon.ai/insights/archive-openai-preparedness-framework-update-april-2025/)
  OpenAI updated its Preparedness Framework on April 15, 2025, revising its severe-harm risk evaluation, safeguards, governance and disclosure approach for frontier capabilities.
- [Microsoft put PyRIT-backed AI red teaming into Foundry preview, making scan evidence a separate release artifact](https://rangoon.ai/insights/archive-microsoft-ai-red-teaming-agent-preview-april-2025/)
  Microsoft announced its AI Red Teaming Agent public preview on April 4, 2025, bringing PyRIT-based adversarial testing to Foundry evaluation, reporting and logging surfaces.
- [The UK retitled its AI Safety Institute and put security risks at the centre of its remit](https://rangoon.ai/insights/archive-uk-ai-security-institute-february-2025/)
  On February 14, 2025, the UK renamed its AI Safety Institute the AI Security Institute, adding a criminal-misuse team and foregrounding national-security research.
- [OECD launched the Hiroshima AI reporting framework, making comparable company disclosures the event—not proof of compliance](https://rangoon.ai/insights/archive-oecd-hiroshima-ai-reporting-framework-february-2025/)
  OECD launched a Hiroshima AI Process reporting framework on February 7, 2025, inviting advanced-AI organizations to disclose risk-management practices in a shared format.
- [FDA’s January AI-device lifecycle draft asked sponsors for an evidence trail, while remaining nonbinding and open for comment](https://rangoon.ai/insights/archive-fda-ai-device-lifecycle-draft-guidance-january-2025/)
  FDA’s January 7, 2025 draft proposed lifecycle and submission documentation for AI-enabled devices, but it was nonbinding draft guidance rather than a new enforceable rule.
- [Genie 3 previewed interactive worlds with a short-horizon evidence boundary](https://rangoon.ai/insights/archive-genie3-interactive-world-model-preview/)
  Google DeepMind announced Genie 3 in August 2025 as a limited research preview for real-time interactive world generation, while documenting duration and accuracy limitations.


## More information

- [Documentation index](https://rangoon.ai/llms.txt)
- [AI access and policies](https://rangoon.ai/ai/)
- [Source repository](https://github.com/hypler-dev/rangoon)