Ex-Anthropic Researcher Quits, Says AI Race Is “Gambling With Our Lives”
Quick Reads
- Jacob Coxon says he resigned from Anthropic after three years of pretraining research at OpenAI and Anthropic.
- He argues that AI companies are moving too quickly toward self-improving superintelligence.
- Coxon is calling for greater coordination and stronger safety measures before AI capabilities advance further.
Former Anthropic researcher Jacob Coxon is raising a serious warning about where the artificial intelligence race could be heading.
In a post on X, Coxon said he spent the past three years working on pretraining research at both OpenAI and Anthropic. He accused both companies of racing toward self-improving superintelligence and described the competition as a dangerous gamble.
Coxon warned that future AI systems could become superhuman, potentially becoming capable of performing complex tasks across areas such as cybersecurity and scientific research at a speed humans cannot match.
He also claimed that people working on advanced AI privately take the possibility of catastrophic outcomes very seriously, even when public statements from executives and researchers sound more cautious. These are Coxon’s personal claims and concerns, rather than established predictions.
One of His Biggest Concerns
One of his biggest concerns is what he calls the AI “endgame” a stage where companies begin developing increasingly powerful systems before researchers have a sufficiently strong understanding of how those systems behave.
Coxon also explained why he believes companies continue building AI despite these risks. He argued that some researchers may not fully appreciate the stakes, while others understand the risks but fear that another company or country could move ahead if they slow down.
Despite his criticism, Coxon said he remains optimistic about coordination between AI companies and governments. He suggested that stronger agreements could help slow the race and, in extreme circumstances, even temporarily limit efforts to increase model capabilities.
His message to AI researchers is clear: Don’t keep racing simply because everyone else is racing. Stop and ask whether the conditions are safe enough to continue.
The AI race is getting faster and Coxon wants the industry to think carefully about where the finish line could lead.
