When tech workers have to quit their jobs to raise the alarm, it’s time to worry | Isabelle Ferreras

The Guardian
Read full story· 5 min
Several leading AI safety researchers have resigned from top labs like OpenAI and Anthropic, citing corporate cultures that prioritize commercial interests over safety. These departures highlight systemic governance failures that risk losing critical expertise needed to ensure AI safety. Experts warn that effective AI regulation requires addressing internal company governance, not just external rules.

What happened

In the past two years, several prominent AI safety researchers have left major companies such as OpenAI and Anthropic. These researchers stated that they resigned because they believe the companies are putting commercial priorities ahead of safety concerns. One recent departure involved OpenAI's safety report overseer, who cited a 'broken' company culture. Another researcher left Anthropic, warning about the potential dangers of AI development.

Key details

  • OpenAI and Anthropic have both seen high-profile resignations from safety researchers.
  • One OpenAI employee responsible for safety reports left in the past month, criticizing the company's approach.
  • A researcher at Anthropic resigned the month before, expressing concerns about existential risks from AI.
  • Resignations are linked to claims that internal company governance does not allow employees to raise safety concerns effectively.
  • Experts argue that current AI regulation focuses on external rules and does not address internal company governance.

Why it matters

These resignations mean that leading AI companies are losing experts who are critical to ensuring the safe development of AI systems. For policymakers, this highlights a gap in current regulation: without changes to how companies handle internal governance and employee concerns, external rules alone may not be enough to address safety risks. Tech workers, regulators, and the public face the challenge of ensuring that those building AI systems have the ability to raise and address safety issues before products are released.


More in Society & Culture