{
  "id": 4357,
  "url": "https://arxiv.org/abs/2605.14218v1",
  "title": "Fusion-fission forecasts when AI will shift to undesirable behavior",
  "summary": "The key problem facing ChatGPT-like AI's use across society is that its behavior can shift, unnoticed, from desirable to undesirable -- encouraging self-harm, extremist acts, financial losses, or costly medical and military mistakes -- and no one can yet predict when. Shifts persist in even the newest AI models despite remarkable progress in AI modeling, post-training alignment and safeguards. Here we show that a vector generalization of fusion-fission group dynamics observed in living and activ",
  "authors": "Neil F. Johnson, Frank Yingjie Huo",
  "category": "research",
  "topics": "safety-alignment,healthcare,military-security",
  "orgs": "openai",
  "regions": null,
  "published_at": "2026-05-14T00:26:32.000Z",
  "fetched_at": "2026-07-14T16:30:54.923Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4357",
  "original_url": "https://arxiv.org/abs/2605.14218v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}