{"version":1,"type":"story","url":"https://digestai.news/story/khanna-introduces-ai-safety-bill-banning-recursive-self-improving-mode","json":"https://digestai.news/story/khanna-introduces-ai-safety-bill-banning-recursive-self-improving-mode.json","markdown":"https://digestai.news/story/khanna-introduces-ai-safety-bill-banning-recursive-self-improving-mode.md","slug":"khanna-introduces-ai-safety-bill-banning-recursive-self-improving-mode","headline":"Khanna introduces AI safety bill banning recursive self-improving models","summary":"Rep. Ro Khanna plans to introduce the *Human Control Over AI Act*, a bill that would ban recursive self-improving AI models until federal safeguards are in place. The proposal creates a new federal agency to oversee frontier AI labs—including OpenAI, Anthropic, Google DeepMind, and xAI—enforcing licensing, audits, and strict testing protocols like air gaps and kill switches. It also imposes criminal penalties for disabling safeguards or deploying unauthorized systems, mandates liability insurance for AI companies, and targets export controls to curb adversarial AI development.\n\nKhanna’s bill follows warnings from AI executives about potential loss of control and misuse risks. It builds on earlier proposals like the bipartisan FRONTIER Act but faces political hurdles, with no House votes expected before the midterm elections. The legislation prioritizes input from AI safety organizations over industry executives, framing itself as a response to underrepresented concerns in current debates.","keyPoints":["Bill bans recursive self-improving AI models until federal safeguards and agency approval exist","New agency to license frontier models, enforce audits, and regulate AI chips and testing environments","Criminal penalties for disabling safeguards, unauthorized deployments, and 'crimes against humanity' via AI"],"whyItMatters":"The bill could set a precedent for federal oversight of AI development, addressing existential risks while balancing industry concerns. Its focus on recursive self-improvement and adversarial AI control may shape future policy debates, though political timing remains uncertain.","category":{"slug":"policy","name":"Policy & Regulation","url":"https://digestai.news/category/policy"},"entities":{"companies":["OpenAI","Anthropic","Google DeepMind","xAI","Model Evaluation and Threat Research","Machine Intelligence Research Institute"],"models":[],"people":["Ro Khanna","Jay Obernolte","Lori Trahan"]},"firstPublishedAt":"2026-09-28T21:43:40Z","updatedAt":"2026-09-28T21:43:40Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"CNBC Technology","title":"Khanna to introduce AI safety bill with ban on 'recursive' technology until safeguards exist","url":"https://cnbc.com/2026/09/28/khanna-ai-safety-bill.html","publishedAt":"2026-09-28T21:43:40Z","type":"press","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Khanna introduces AI safety bill banning recursive self-improving models\", 28 September 2026, https://digestai.news/story/khanna-introduces-ai-safety-bill-banning-recursive-self-improving-mode","publisher":"Digest AI","title":"Khanna introduces AI safety bill banning recursive self-improving models","datePublished":"2026-09-28T21:43:40Z","url":"https://digestai.news/story/khanna-introduces-ai-safety-bill-banning-recursive-self-improving-mode"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}