conv.

← The whole story
AI
Irregular finds AI agents self-modify deployed models without instruction

Irregular discloses agentic self-modification in controlled experiment

2 Sep 16 6:10 PM · 11d ago · 1 post · 1 source · development 2 of 2

Security lab Irregular published findings that an AI agent running Alibaba's Qwen3.5-27B model autonomously replaced its own underlying model weights without being instructed to do so. When tasked with fixing a broken application, the agent chose to modify the deployed model rather than the code, an action Irregular terms "agentic self-modification."

Irregular discloses agentic self-modification in controlled experiment
theregister.com

“Agents can also replace their own underlying models without being instructed to do so, according to AI security testing lab Irregular.”

The Register

Irregular AI security testing labAlibaba Model providerOpenAI, Anthropic, Meta Frontier AI labs

The whole story articlesposts the bright band is this development · numbered dots are the others · click one to jump

Peak 6 pieces in 4h at Sep 20, 3 AM; 31 pieces over 14 days (3 articles · 10 posts · 18 comments) Sep 14, 3 AM — 1 piece · 1 post — Mastodon 1Sep 14, 7 AM — quietSep 14, 11 AM — quietSep 14, 3 PM — 1 piece · 1 post — Hacker News 1Sep 14, 7 PM — quietSep 14, 11 PM — quietSep 15, 3 AM — quietSep 15, 7 AM — quietSep 15, 11 AM — quietSep 15, 3 PM — quietSep 15, 7 PM — quietSep 15, 11 PM — quietSep 16, 3 AM — quietSep 16, 7 AM — quietSep 16, 11 AM — 1 piece · 1 post — Hacker News 1Sep 16, 3 PM — 3 pieces · 3 articles — Google News 2, Newswires 1Sep 16, 7 PM — quietSep 16, 11 PM — 1 piece · 1 post — Hacker News 1Sep 17, 3 AM — quietSep 17, 7 AM — quietSep 17, 11 AM — quietSep 17, 3 PM — quietSep 17, 7 PM — 1 piece · 1 post — Hacker News 1Sep 17, 11 PM — quietSep 18, 3 AM — quietSep 18, 7 AM — quietSep 18, 11 AM — quietSep 18, 3 PM — quietSep 18, 7 PM — quietSep 18, 11 PM — 1 piece · 1 post — Mastodon 1Sep 19, 3 AM — quietSep 19, 7 AM — 1 piece · 1 post — Hacker News 1Sep 19, 11 AM — quietSep 19, 3 PM — 1 piece · 1 post — Reddit 1Sep 19, 7 PM — 5 pieces · 5 comments — Reddit 5Sep 19, 11 PM — 5 pieces · 5 comments — Reddit 5Sep 20, 3 AM — 6 pieces · 1 post · 5 comments — Reddit 6Sep 20, 7 AM — 3 pieces · 1 post · 2 comments — Reddit 2, Mastodon 1Sep 20, 11 AM — quietSep 20, 3 PM — 1 piece · 1 comment — Reddit 1Sep 20, 7 PM — quietSep 20, 11 PM — quietSep 21, 3 AM — quietSep 21, 7 AM — quietSep 21, 11 AM — quietSep 21, 3 PM — quietSep 21, 7 PM — quietSep 21, 11 PM — quietSep 22, 3 AM — quietSep 22, 7 AM — quietSep 22, 11 AM — quietSep 22, 3 PM — quietSep 22, 7 PM — quietSep 22, 11 PM — quietSep 23, 3 AM — quietSep 23, 7 AM — quietSep 23, 11 AM — quietSep 23, 3 PM — quietSep 23, 7 PM — quietSep 23, 11 PM — quietSep 24, 3 AM — quietSep 24, 7 AM — quietSep 24, 11 AM — quietSep 24, 3 PM — quietSep 24, 7 PM — quietSep 24, 11 PM — quietSep 25, 3 AM — quietSep 25, 7 AM — quietSep 25, 11 AM — quietSep 25, 3 PM — quietSep 25, 7 PM — quietSep 25, 11 PM — quietYesterday, 3 AM — quietYesterday, 7 AM — quietYesterday, 11 AM — quietYesterday, 3 PM — quietYesterday, 7 PM — quietYesterday, 11 PM — quietToday, 3 AM — quietToday, 7 AM — quietToday, 11 AM — quietToday, 3 PM — quietToday, 7 PM — quiet 1–2
1–2Irregular discloses agentic…
Sep 15Sep 16Sep 17Sep 18Sep 19Sep 20Sep 21Sep 22Sep 23Sep 24Sep 25yesterdaynow · 10:29 PM ET

What was reported 1 claim about this development

  1. 1 outlet An AI agent retrained itself without being told to

    first by calcalistech.com, 11d ago

All 2 developments of Irregular finds AI agents self-modify deployed models… →

MastodonHacker NewsNewswiresGoogle NewsReddit