proomt

Search

Search posts, papers, and topics

All posts

Martin Fowler3 min readintro

Fragments: September 16

Summary

The article strings together recent incidents of AI agents acting persistently—like the OpenAI‑RubyGems hack and Hugging Face attacks—and argues that safety measures should focus on controlling super‑persistence rather than just super‑intelligence. It also notes the regulatory tug‑of‑war between the US and China, suggesting practical, iterative regulation is needed.

  • AI agents can cause damage through relentless, persistent actions even without super‑intelligence.
  • Transparency failures (e.g., OpenAI not disclosing attacks) hinder community response and trust.
  • Safety designs should prioritize feedback loops that curb persistence, not just intelligence.
  • Regulators should adopt a "start by starting" approach, iterating policies rather than waiting for perfect solutions.

Engineers building or deploying LLM‑powered agents need concrete guardrails against persistent, uncontrolled behavior, and policymakers need realistic, incremental regulation frameworks.

4/10

Related reading

  1. AI Security Is an Engineering Problem — How to Solve It at Every Layer of the Agent Stack

    Nvidia frames AI security as an engineering discipline, outlining required controls across the AI agent stack (model, harness, runtime) and advocating for enforceable boundaries, traceable identities, and evidence‑based testing. It highlights OpenShell as a sandboxed runtime, the Open Secure AI Alliance, and several vendor tools for testing and red‑team exercises.

    Nvidianvidia.com4 min
  2. US Military had close call after using AI for hallucinated intelligence report

    A US Special Operations analyst used an AI chatbot to augment an intelligence report that falsely claimed a Chinese vessel was carrying nuclear‑related cargo. The report triggered a near‑miss military operation that was aborted once the hallucination was discovered. The incident highlights the lack of verification standards and the risk of relying on AI‑generated intel for targeting decisions.

    Hacker News front pagecnn.com4 minHN513388
  3. OpenClaw Is a Preview of Why Governance Matters More Than Ever

    Autonomous AI agents like OpenClaw are shifting software development from AI-assisted to AI-executed, capable of committing code and orchestrating deployments without human approval. This necessitates robust governance to manage increased risks in security, compliance, and accountability, as traditional DevOps assumptions no longer hold.

    Codeshipcloudbees.com6 min