# Former Anthropic Employees Describe Debate Over AI Safety and Competition

By Simon Yoon

Canonical URL: https://www.tokenpost.com/news/technology/61298
Published: 2026-11-19T12:13:30.000Z
Updated: 2026-11-19T12:13:30.000Z
Section: Technology

> Anonymous former employees recounted CEO Dario Amodei’s changing view of the AI race and internal disagreements over whether to slow development.

Anonymous former Anthropic employees described disagreements over the company’s AI safety mission and competition, including CEO Dario Amodei’s changing view of whether the race to build more powerful systems can be stopped.

One former employee recounted Amodei discussing the possibility that Anthropic could someday become more powerful than governments and need relationships with countries. Another compared superintelligence to the One Ring from “The Lord of the Rings,” saying the first group to develop it could gain dangerous cyber weapons, ways to cure diseases and an inexhaustible AI workforce. The employee said concentrating that power among a small group would make trust in its holders crucial.

In 2022, Amodei said Anthropic should not race ahead or train models far larger than competitors’ models. That year, the company delayed offering Claude to the public over safety concerns, while ChatGPT gave OpenAI a market head start.

Amodei now believes the AI race cannot be stopped: if Anthropic does not develop the most advanced models, other companies will continue. He believes staying ahead would give Anthropic enough influence to encourage safer practices across the industry.

The safety mission also shapes Anthropic’s hiring. A famous interview question asks candidates whether they would still work at the company if slowing research for AI safety made their stock holdings worthless. Research teams would reject technically excellent candidates who did not agree with the company’s philosophy.

As Anthropic has expanded, some early employees worried that new sales and marketing staff may have joined for high salaries and stock without believing in the safety mission. Other employees urged the company to stop training more powerful AI systems even if competitors such as OpenAI kept going. Amodei rejected that proposal.


