Steven Adler is an AI safety researcher, a co-founder of Guidelight AI Standards, and the author of the Clear-Eyed AI newsletter. He spent four years at OpenAI, where by his own account he led safety-related research for "first-time product launches" and "more speculative long-term AI systems," and left the company in November 2024 (Source: theguardian.com).
Departure from OpenAI and public position
Adler set out his reasons for concern in a series of posts on X in late January 2025, describing his time at OpenAI as "a wild ride" while saying he is "pretty terrified by the pace of AI development these days" and asking, in the context of raising a family or saving for retirement, "will humanity even make it to that point?" He characterized the pursuit of artificial general intelligence as "a very risky gamble, with huge downside," argued that no research lab has a solution to alignment, and warned that competitive pressure makes one less likely to be found: "The faster we race, the less likely that anyone finds one in time." His prescription was regulatory rather than voluntary — the industry, he said, is "stuck in a really bad equilibrium" and needs "real safety regs," because "even if a lab truly wants to develop AGI responsibly, others can still cut corners to catch up, maybe disastrously" (Source: theguardian.com).
The posts appeared in the same week that DeepSeek's release of a model competitive with OpenAI's at apparently lower cost unsettled the US industry, which sharpened the competitive-pressure argument he was making (Source: theguardian.com).
Guidelight AI Standards
Through Guidelight, Adler published the Control standard v1.0 in May 2026, six minimum principles for preventing loss of control during internal deployment of advanced AI systems, and announced in July 2026 that Guidelight was conducting its first assessment of leading AI companies against the standard (Source: clear-eyed.ai). The work extends the AI control agenda — deployment safeguards that remain safe even if a model is misaligned — into third-party standard-setting, and is one instance of the broader turn toward external assurance bodies also visible in the independent-verification-organization proposals of Americans for Responsible Innovation.
Relationships
- related: Guidelight AI Standards (co-founder), AI Control, OpenAI (former employer)
- supports: AI Race Dynamics — argues competitive pressure prevents alignment work from completing