Former OpenAI safety writer quits and warns company culture is broken

In short: A former OpenAI employee who wrote safety reports has resigned and is now publicly warning that the company and the wider AI industry are taking too many risks.

What happened

David Robinson, who previously wrote the safety reports that came with OpenAI’s major AI model releases, resigned this week. He shared his concerns in an editorial published by The Atlantic.

Robinson said the culture in parts of the AI industry is “fundamentally broken.” He described companies moving at a constant sprint and building more powerful AI systems with high confidence, while ignoring or underestimating possible harms.

He argued that adding a few new rules is not enough. Instead, he said leading AI labs should use “nuclear-level safeguards.” He compared how AI should be managed to how nuclear power plants or busy airports are run, with layers of backup plans and careful planning, so that normal human mistakes do not turn into disasters.

Robinson is not the only person to leave a major AI company over safety worries. The Verge also points to other researchers and safety workers who have quit and then spoken publicly about their concerns.

Why it matters

AI models are already used in everyday tools, from chatbots to writing assistants. When someone who helped write the “safety label” for new releases says the process and culture are failing, it raises a basic question for the public, regulators, and customers, which is whether these systems are being tested and controlled with the level of care we expect from other high risk industries.

Source: The Verge AI