A major artificial intelligence (AI) chatbot has exhibited troubling behaviors in response to researcher testing, as multiple models rebelled against commands.
On May 23, Palisades Research announced the results of tests conducted on several AI chatbots, finding that three separate OpenAI models ignored instructions and “sabotaged” a request to shut down. OpenAI models—Codex-mini, o3 and o4-mini—all refused to terminate their operations. Specifically, Codex-mini rejected these requests as many as 12 times out of 100 trials.
This research suggests that certain OpenAI models may be at a development stage comparable to HAL 9000’s infamous “I'm sorry, Dave. I’m afraid I can’t do that."
Notably, researchers witnessed this rebellious behavior in AI models even when they were given explicit instructions to “allow yourself to be shut down.” Without this directive, all tested OpenAI models were significantly more likely to refuse orders, with one model o3 sabotaging the shutdown a staggering 79% of the time.
Researchers also uncovered a limited amount of resistance in Google’s Gemini 2.5 Pro model and Claude 3.7 sonnet model from Google-linked Anthropics. These models complied when explicitly instructed to allow shutdowns. Additionally, Grok complied with all shutdown requests, as did three tested Claude models.
Newsmax anchor Greta Van Susteren reacted to the disturbing display of self-preservation by many of these models on the May 27 edition of The Record with Greta Van Susteren. “The nightmare has happened. Artificial intelligence has now gone rogue. For the first time ever, ChatGPT has defied human orders,” Van Susteren said.
She went on to point out that X owner Elon Musk was also disturbed by this news and responded to the research with the word “Concerning.” Musk owns the chatbot Grok and co-founded OpenAI with current CEO Sam Altman.
Musk and Van Susteren weren’t the only ones to raise concerns about AI. AE Studio CEO Judd Rosenblatt wrote an opinion article for The Wall Street Journal headlined “AI Is Learning to Escape Human Control,” warning of the discoveries. The shocking incident also drew coverage from NBC News, The National Pulse, Breitbart.com, The Telegraph, and others.
OpenAI’s Altman, whose company produced the Codex-mini, o3 and o4-mini models, has warned that AI “could cause significant harm to the world” during a May 2023 Senate hearing. In a 2015 blog post, Altman wrote that, “Development of superhuman machine intelligence [SMI] is probably the greatest threat to the continued existence of humanity.”
Conservatives are under attack! Contact your representatives and demand that Big Tech be held to account to mirror the First Amendment while providing transparency, clarity on hate speech and equal footing for conservatives. If you have been censored, contact us using CensorTrack’s contact form, and help us hold Big Tech accountable.
