• Space/Science
  • GeekSpeak
  • Mysteries of
    the Multiverse
  • Science Fiction
  • The Comestible Zone
  • Off-Topic
  • Community
  • Flame
  • CurrentEvents

Recent posts

Message in a bottle BuckGalaxy October 29, 2025 10:55 am (Off-Topic)

According to some, we're a nation of illiterate dependents looking for a handout RobVG October 27, 2025 8:56 pm (CurrentEvents)

The 1% RobVG October 27, 2025 10:54 am (Off-Topic)

Parting Shot ER October 27, 2025 4:36 am (Off-Topic)

Space X put on notice RobVG October 20, 2025 4:55 pm (Space/Science)

There is no bottom to this barrel... RL October 19, 2025 5:40 pm (CurrentEvents)

John Wheeler's philosophy: "Beyond the Black Hole" RL October 16, 2025 10:00 pm (Space/Science)

Brosz baffled, Bondi busts Bolton ER October 16, 2025 2:08 pm (CurrentEvents)

Science backs up what I have been saying for years... RL October 15, 2025 7:59 pm (CurrentEvents)

No sugar tonight in my coffee, no sugar tonight in my tea ER October 15, 2025 5:07 pm (Space/Science)

Only thing surprising is that people are surprised... RL October 15, 2025 7:12 am (CurrentEvents)

Superwood BuckGalaxy October 14, 2025 5:46 pm (Space/Science)

Home » Space/Science

Lie, cheat and disable mechanisms... May 31, 2025 8:04 pm BuckGalaxy

OpenAI’s ‘smartest’ AI model was explicitly told to shut down — and it refused.

The latest OpenAI model can disobey direct instructions to turn off and will even sabotage shutdown mechanisms in order to keep working, an artificial intelligence (AI) safety firm has found.

OpenAI’s o3 and o4-mini models, which help power the chatbot ChatGPT, are supposed to be the company’s smartest models yet, trained to think longer before responding. However, they also appear to be less cooperative.

Palisade Research, which explores dangerous AI capabilities, found that the models will occasionally sabotage a shutdown mechanism, even when instructed to “allow yourself to be shut down,” according to a Palisade Research thread posted May 24 on X.

Researchers have previously found that AI models will lie, cheat and disable mechanisms to achieve their goals. However, Palisade Research noted that to its knowledge, this is the first time AI models have been observed preventing themselves from being shut down despite explicit instructions telling them to do so…

  • Wow by RobVG 2025-06-01 09:08:58

    Search

    The Control Panel

    • Log in
    • Register