A former researcher at Anthropic has warned about the dangers of the AI development race, stating that "the possibility of humanity losing control over AI is greater."
On October 5, Jacob Coxon, a former researcher at Anthropic, testified at a New York City Council hearing on AI, saying, "We still do not know how to control any AI system."
Coxon worked on pretraining research at OpenAI and Anthropic for three years. He announced his departure on social media on September 8, criticizing both companies for acting irresponsibly, stating, "Neither company is acting responsibly. They are gambling with our lives."
During the hearing, he added, "If we continue on the current path, the likelihood of humanity losing control over AI is greater than the likelihood of not losing control, and this could lead to human extinction."
Coxon was invited to the hearing by City Council Chair Julie Menin, who is pushing for a package of AI regulation bills. This package consists of ten ordinances, including a mandate for a 'kill switch' to immediately halt systems in emergencies, a requirement for third-party safety verification before AI systems can be sold or distributed in New York City, the introduction of a whistleblower compensation program, and fines of up to $25,000 for regulatory violations.
Coxon noted that while the 'kill switch' could be helpful in the short term, it is not sufficient to ensure long-term safety.
Representatives from major tech companies, including OpenAI, Anthropic, and Google, also testified via video at the hearing. The City Council stated it would seek a subpoena for SpaceX's AI division, which did not respond to the summons.
As the Trump administration generally maintains a hands-off approach to AI regulation, state and local governments like New York are accelerating their independent regulatory efforts, leading to increasing discrepancies between federal and local policies.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.