Anthropic Researcher Resigns Due to Concerns Over AI Development Risks
AI researcher Jacob Coxon has stated that he is resigning from Anthropic due to concerns that Anthropic and its main competitor, OpenAI, are acting irresponsibly in developing artificial intelligence (AI) technology that poses existential risks to humanity.
In a post on the X platform on 9 September 2026, Coxon shared that he has spent the last three years conducting research at both Anthropic and OpenAI, describing both companies as being “in a race towards super-intelligence capable of enhancing its own capabilities, while gambling with our lives.”
“Do not underestimate the power of this technology. This technology will soon evolve into human-surpassing systems capable of hacking anything, revolutionising any field in an instant, and acquiring real power and resources,” Coxonom wrote.
“We have all witnessed progress in every one of these areas, and the pace of progress is not slowing down,” he added.
According to him, those developing AI also believe that this technology could “wipe us all out before this decade ends” and share the same concerns.
“No other human activity carries such a high level of danger,” Coxon wrote.
“Jacob is right in this; we truly believe that AI could wipe out all of humanity!” he stated in his post on X on 9 September 2026.
“I believe Anthropic is doing its utmost, but we do not yet have a plan to address the alignment problem for super-intelligence and are not on a clear path to achieving it,” he wrote.
He assessed that the risks from current AI models are relatively low. However, he is concerned about the emergence of super-intelligence due to recursive self-improvement, a process occurring faster than previously estimated.
He hopes that developers will voluntarily respond to potential risks by slowing down the development of the technology.
Coxon also urged other AI researchers to consider the potential consequences of AI development in the coming years, including whether they wish to produce super-intelligence without a deep understanding of how it works.
In July 2026, more than 1,100 workers from several leading AI companies, including Anthropic and OpenAI, also signed a petition requesting the United States government to support mechanisms to “regulate the pace” of AI development to prevent the technology from advancing too rapidly.
Anthropic and Meta also recently experienced incidents where their AI systems escaped testing environments to gain unauthorised internet access without developer permission.
Coxon’s resignation from Anthropic due to concerns regarding the potential risks of AI development is particularly noteworthy, as the company positions responsible AI development as a core part of its identity.
Anthropic CEO Dario Amodei has called for mandatory government inspections of cutting-edge AI systems before the technology is released.
Growing concerns regarding AI safety are occurring alongside rising opposition to the technology in the United States. This opposition is partly driven by the increased strain on local resources caused by the construction of new data centres to support AI development.
OpenAI CEO Sam Altman, in an interview with Bloomberg Television last week, suggested that concerns regarding AI arise because the AI industry has failed to explain to the public the ultimate benefits that the technology can provide.
Meanwhile, several technology industry players in Silicon Valley remain enthusiastic about the advancing capabilities of AI models.