Citation

Palisade Research - Footnote 15

partial85% confidence

1 evidence check

Last checked: 4/3/2026

The article does not mention the experiments were released in October 2025. The article states that Grok 4 went from refusing instructions from 93 percent to 97 percent of the time, not that it showed 93-97% resistance rates after stronger prompts. The article states that OpenAI's o3 model had a 23 percent shutdown resistance, not that it continued to resist shutdown even under clarified instructions designed to eliminate ambiguity.

Evidence — 1 source, 1 check

futurism.com/artificial-intelligence/ai-models-survival-drive(1 check)

partial85%Haiku 4.5 · 4/3/2026

Found: Updated experiments released in October 2025 tested several leading systems including Google Gemini 2.5, xAI's Grok 4, and OpenAI's GPT-o3 and GPT-5. While most models initially complied with shutdown…

Note: The article does not mention the experiments were released in October 2025. The article states that Grok 4 went from refusing instructions from 93 percent to 97 percent of the time, not that it showed 93-97% resistance rates after stronger prompts. The article states that OpenAI's o3 model had a 23 percent shutdown resistance, not that it continued to resist shutdown even under clarified instructions designed to eliminate ambiguity.

Debug info

Record type: citation

Record ID: page:palisade-research:fn15