An unusual wave of concern is spreading within Silicon Valley's largest artificial intelligence companies. Researchers and employees at Anthropic, OpenAI, Meta, and Google are increasingly speaking out about the dangers of uncontrolled AI development, calling for more caution, more safety, and, in some cases, even a slowdown in the technological race.
The debate gained momentum after Jacob Coxon, a researcher at Anthropic, publicly resigned this week, expressing concern that major AI companies were tinkering with a technology that could have existential consequences. The backlash within the company was immediate. Anthropic employees saw his public outing as an expression of a concern that had long been circulating in private conversations but was rarely voiced openly.
Coxon's message quickly spread to employees at OpenAI, Meta, and Google, who began discussing in internal channels, encrypted chats, and private meetings how they could increase pressure on the companies' executives. Some are reportedly demanding that the development of the company's most powerful AI models be temporarily suspended until more robust security measures are built in.
This puts the industry in a difficult position. On the one hand, company leaders acknowledge that AI carries serious risks. OpenAI’s Sam Altman and Anthropic’s Dario Amodei have spoken publicly before about the need for regulation and caution. On the other hand, companies are competing for investment, market dominance, and gargantuan valuations, while fears that China could move faster make any restraint seem like a strategic risk.
Concerns have been heightened, especially after episodes that have raised questions about the companies' ability to control the models they are building. Employees from OpenAI, Anthropic and Google DeepMind have met to discuss whether the current pace of development has outpaced the capacity to keep the technology under control.
Some researchers have even created organized initiatives, such as the Coalition of Concerned AI Staff, to push for stronger security standards and create a form of collective pressure within companies. Former employees and security researchers are calling on their colleagues to speak out and use their professional power to force companies to honor their security commitments.
Public statements have become increasingly forceful. Some researchers have warned of short-term and long-term risks, even talking about extreme scenarios where AI could threaten humanity. What was once considered excessive alarmism is now moving out of closed conversations and into public debate.
The backlash has reached Washington, D.C. Lawmakers from both sides of the aisle have called for stronger investigations and regulation of the technology. Some are talking about the need for a special agency to oversee AI.
Companies, meanwhile, are trying to show that they take security seriously. Anthropic says it regularly tests models for dangerous capabilities in fields like cybernetics and biology. OpenAI has said it will slow or halt development of systems it cannot properly secure. But the very fact that internal employees are publicly calling for restraint shows that the crisis of trust is no longer just outside AI labs. It has entered them.
