OpenAI Recruits Anthropic AI Safety Veteran Amid Senior Staff Departures
As internal projects struggle to compete with ChatGPT for computing credits and approvals, researchers question the future of long-term work at OpenAI.
News
- Nvidia CEO Pushes Back Fear of AI Replacing Software Tools
- OpenAI Recruits Anthropic AI Safety Veteran Amid Senior Staff Departures
- Waymo Secures $16 Billion to Expand Robotaxi Operations Worldwide
- OpenAI, Snowflake in $200 Million Deal to Embed AI in Enterprise Data
- Musk Merges SpaceX and xAI in Deal Valued at $1.25 Trillion
- TASC Group Appoints New CEO To Lead AI-Driven Growth in the MENA Region
OpenAI has appointed Dylan Scandinaro, formerly of rival AI firm Anthropic, as its new Head of Preparedness, filling a role that drew attention late last year for its unusually high compensation and its significance amid concerns around AI safety and governance.
The appointment was announced by OpenAI chief Sam Altman in a post on X. Scandinaro previously worked on AI safety at Anthropic, a San Francisco–based start-up that has positioned itself as a more risk-conscious alternative to OpenAI.
In his new role, Scandinaro will be responsible for ensuring the safe development and deployment of increasingly powerful AI systems, as well as preparing the company for the risks associated with their use.
The hiring underscores both OpenAI’s growing emphasis on operational safety and the escalating competition for talent among leading AI developers. Anthropic has gained industry recognition for its work on guardrails and its so-called “AI Constitution,” an approach designed to constrain model behaviour through explicit principles. That focus has supported its enterprise adoption, according to analysts.
Altman acknowledged the scale of the challenge ahead, noting that OpenAI expects to work with “extremely powerful models” in the near future. Scandinaro, he said, “has his work cut out for him.”
The move comes as OpenAI undergoes a broader strategic shift, prioritising rapid improvements to ChatGPT over longer-horizon experimental research. According to Financial Times sources, multiple current and former employees, the company has reallocated resources away from exploratory projects in favour of advances to the large language models that power its flagship ChatGPT.
That shift has coincided with a series of high-profile departures. Among those to leave in recent months are vice-president of research Jerry Tworek, model policy researcher Andrea Vallone, and economist Tom Cunningham. Several of those departures were linked, directly or indirectly, to disagreements over resource allocation and research priorities.
OpenAI emerged as a research lab, with ChatGPT launched in 2022 as a research preview before catalysing the generative AI boom. Now, under pressure to justify a valuation of roughly $500 billion, the company is evolving into one of Silicon Valley’s largest product-driven AI platforms. That transition has sharpened internal debates over whether OpenAI still offers space for what some researchers describe as “blue-sky” work.
People familiar with the company say researchers must increasingly compete for computing credits and executive approval to pursue projects. Teams working on non-language-model initiatives, including video and image generation systems such as Sora and DALL·E, were seen as under-resourced as ChatGPT became the central organisational priority.
Mark Chen, OpenAI’s chief research officer, has pushed back against the idea that long-term research is being sidelined. He said foundational research continues to account for the majority of the company’s compute and investment, with hundreds of projects exploring long-horizon questions. Pairing research with large-scale deployment, Chen argued, strengthens scientific rigour by accelerating feedback and learning loops.
Nonetheless, former employees describe an increasingly competitive internal environment shaped by external pressures. Google’s Gemini models and Anthropic’s Claude system have both made gains in benchmarks and code generation, intensifying the pace of development across the sector.
Some investors remain unconcerned about the risk that OpenAI will fall behind in model quality alone. Jenny Xiao, a partner at Leonis Capital and a former OpenAI researcher, argues that the company’s advantage lies less in marginal technical superiority than in its vast user base.
“The moat has shifted from research to user behaviour,” she said. “That’s a much stickier advantage.”
