OpenAI Hires Anthropic's Dylan Scandinaro as Head of Preparedness

OpenAI, the company behind ChatGPT, has announced the appointment of Dylan Scandinaro as its new Head of Preparedness — a strategic move to strengthen its team focused on advanced AI safety and risk assessment. Notably, Scandinaro joins from a role focused on AGI safety at Anthropic, OpenAI's fiercest rival in Silicon Valley.

A Warning and Optimism in Equal Measure

In his first statement following the appointment, Scandinaro wrote on X (formerly Twitter): "AI is advancing at an astonishing pace. The potential benefits are enormous, as are the risks of severe or even irreversible harm. There is a lot of work to do, and not a lot of time to do it!"

This statement reflects the tone of "cautious urgency" that has come to dominate among safety researchers in 2026, as models grow more powerful at an accelerating rate.

The AI Safety Talent War

This hire is being read across the tech industry as part of the ongoing "talent war" between AI giants. Anthropic, originally founded by former OpenAI researchers (the Amodei siblings) over safety concerns, has long been considered the top destination for researchers who prioritize safety over speed. A prominent figure like Scandinaro moving to OpenAI could signal:

1. OpenAI's growing seriousness about integrating stricter safety standards.
2. OpenAI's formidable ability to attract top talent from competitors, thanks to its resources and influence.

Are We Close to AGI?

This news coincides with recent controversial media reports. Forbes quoted Sam Altman, OpenAI's CEO, with a vague hint that the company has "basically built AGI, or something very close to it." Although Altman later clarified the statement was "spiritual" rather than technically literal, the appointment of a new Head of Preparedness — the team responsible for anticipating catastrophic risks from superintelligent models — reinforces the belief that the company is approaching an entirely new technological threshold.

What's Next?

Scandinaro will be responsible for leading OpenAI's efforts to assess future risks and ensure that upcoming models (possibly GPT-6 and beyond) do not pose threats to cybersecurity or human safety. The open question remains: can OpenAI balance its aggressive commercial ambitions with the warnings of its new safety team?