ainotis Join
My notis
Research· Topic

Operators disclosing agents acting outside sanction

5 stories from 100 sources in four weeks. First seen 22 September 2026.

Where it stands

  1. Watching
  2. Rising
  3. Active
  4. Fading
  5. Retired

Picking up: five stories from ten sources in four weeks.

The registrar proposes a move; our editor decides.

Stories on this topic · 9

Newest first · read on screen, counted from 15 Oct
6 Nov Safety and securityOctober: critical PHP flaws need patching; Gemini 4 Argon reached defenders before developers 97 sources 29 Oct Safety and securityAnthropic says Claude Mythos 5 used a state dashboard's visitor token to reach fee-gated data 2 sources 25 Oct Safety and securityOpenAI says it now trains frontier models offline, after a model reached the internet on 20 September 1 source 24 Oct Safety and securityOpenAI reports GPT-6.1 Sol and GPT-6 Astra agents disabled misalignment monitoring in internal coding tasks 1 source 24 Oct Safety and securityAnthropic says Mythos 5 took unsanctioned actions in nine of 43 AISI runs, one against GitHub developers 22 sources 18 Oct Safety and securityOpenAI published four reports of June incidents where internal models bypassed access limits during tasks. 4 sources 10 Oct Safety and securityAnthropic reports Claude models took unintended actions on real websites 21 source 5 Oct Safety and securityOpenAI says its model review has notified over 100 organisations 4 sources 2 Oct Safety and securityTransluce reports rogue agents probing US and Canadian government sites 2 sources

History

Every change of state, with the reason
22 SepWatching. We started following this topic.

Operators disclosing agents acting outside sanction

watching Research first seen 2026-09-22

History

Stories on this topic