Anthropic researcher resigns, warns AI could kill humanity

Jacob Kossen, a researcher who worked at OpenAI and Anthropic, has resigned from Anthropic and issued a stark warning about the future of artificial intelligence.

In a post on the “X” platform, Kossen said he had conducted research into pretraining at both companies over the past three years. He wrote that the companies were acting irresponsibly and “gambling with human lives by racing toward self-improving superintelligence.”

According to Kossen, future AI systems could surpass humans in various fields, create complex cyberthreats and eventually develop their capabilities independently. He said people working on AI seriously believed the technology could destroy all of humanity by the end of the decade.

Evan Hubinger, who works at Anthropic, said Kossen’s views were accurate. He said he estimated the probability that AI would wipe out all humans within the next 10 years at more than 10 percent.

Hubinger stressed that Anthropic was doing everything it could to address the problem, but that there was still no plan for aligning superintelligence with human values.

Anna Wang, who previously worked at Google DeepMind and now works at Anthropic, said the concern was widespread among her colleagues. She said there was still no workable scientific plan to eliminate the risks posed by AI that could repeatedly improve itself.

Alex Turner, a former Google DeepMind researcher, also backed Kossen’s views. He said many researchers believed they were creating a technology that could cause the deaths of everyone on the planet.

The warning also drew the attention of US lawmakers. US Senator Bernie Sanders backed Kossen and said he would introduce legislation to ban artificial superintelligence and pause AI development.

Republican Senator Ted Cruz described the possible consequences of AI as a “catastrophic risk” and called for stronger safety measures.

Anthropic said it had always been clear that AI could create both major benefits and unprecedented risks. The company said it was working on safety mechanisms for deploying more powerful models and supported broader coordination.

Kossen believes voluntary restrictions will not be sufficient. In his view, as long as competition between companies continues, researchers should demand different conditions instead of remaining silent and saying, “this is happening anyway.”