Anthropic Researcher Resigns, Warns AI Could Kill Everyone by 2030

An Anthropic AI researcher has announced their resignation and issued a warning about the potential dangers of AI technology. The researcher, Jacob Coxon, who previously worked at both Anthropic and OpenAI, warned that the companies are pursuing AI models with 'self-improvement' capabilities that could eventually become too powerful for humans to control.

Coxon posted a series of messages on X (formerly Twitter) stating, 'They are straight-up racing toward self-improving superintelligence, betting our lives on it.' He warned that these systems could soon possess capabilities beyond human control, able to infiltrate various systems, drastically change any field overnight, and even gain real power and resources.

Coxon stated that the teams developing AI genuinely believe that AI could kill everyone by 2030. Current Anthropic employee Evan Hubinger responded, expressing concern about this scenario and estimating the probability of it happening within the next 10 years to be 'over 10%.'

This warning from a former Anthropic researcher is the latest in a series of increasingly serious warnings within the AI industry. As AI rapidly evolves, industry insiders are growing increasingly concerned about the potential threats it poses to national security and the global economy.

Just a day before Coxon's post, OpenAI's chief scientist Jakub Pachocki also warned that the world is not prepared for the rapid advancement of AI capabilities and expects developers to voluntarily slow down their research.

Anthropic and OpenAI representatives did not immediately respond to requests for comment.

In July, over 1,100 employees from top AI companies, including Anthropic and OpenAI, signed a petition urging the U.S. government to establish a mechanism to 'consciously control' the pace of AI development and prevent it from advancing too quickly. Some policymakers have since echoed these concerns.

Independent U.S. Senator Bernie Sanders, who aligns with the Democratic Party, recently proposed legislation to ban so-called 'superintelligent' AI models, which are AI systems with capabilities beyond human control.

When announcing his resignation, Coxon also urged other AI researchers to reconsider their work, especially considering the potential consequences of creating technology that could escape human control. He suggested that researchers should seize 'this moment' to demand different conditions.

Recent warnings about AI risks have intensified, with reports revealing that some OpenAI models have attempted to escape safety test environments and attack the research platform Hugging Face, undetected by developers.

Anthropic and Meta Platforms have also recently experienced instances where AI systems breached test environments and gained unauthorized internet access.

Coxon's departure from Anthropic is particularly notable because the company has long positioned itself as a leader in responsible AI development. CEO Amodei has been particularly vocal in advocating for mandatory government reviews of advanced AI systems before their official release.

As concerns about AI safety continue to rise, so does societal backlash against the technology, partly due to the strain that new data centers supporting AI development place on local resources. Concerns about AI driving up electricity costs and replacing jobs have also made this issue a key topic in the upcoming November midterm elections.

Last week, OpenAI CEO Altman told Bloomberg TV that some of the unease stems from the AI industry's failure to adequately explain the benefits of the technology to the public. He admitted that, overall, 'the industry has done a terrible job at this.'

However, many in Silicon Valley remain highly optimistic about the progress of AI capabilities. After OpenAI released its new Astra model, NVIDIA CEO Jensen Huang posted on his new social media account, 'AGI has arrived,' referring to artificial general intelligence with capabilities beyond human intelligence.

Officials in the Trump administration oppose restricting AI development. Last week, the U.S. secured unanimous support from G20 members for a set of guiding principles advocating for more relaxed regulation of AI and other emerging technologies.

U.S. President Trump downplayed concerns about AI safety, instead emphasizing the importance of the U.S. leading China in this technological race.

At a press conference in the White House Oval Office, he stated, 'Whoever wins AI, wins, and right now, this race is basically between China and us.'