Generous Compute & Funding
OpenAI is making a significant investment in AI safety research through a new fellowship program. A core component of this initiative is the allocation
of substantial computational power, with fellows set to receive up to $15,000 in AI compute resources each month. This compute access is a critical element of their sales pitch, aimed at attracting top talent to work closely with OpenAI's frontier lab safety team. In the broader AI landscape, compute power has become a key indicator of prestige and capability among leading technology firms. For instance, Jensen Huang, CEO of Nvidia, recently emphasized the importance of AI engineers utilizing significant compute resources, equating it to a substantial financial investment. The fellowship period is scheduled from September 14, 2026, to February 5, 2027. Participants will receive a weekly stipend of $3,850, which translates to an annual equivalent salary exceeding $200,000 before taxes, not accounting for holidays. In total, fellows are expected to earn over $111,000 during the program's duration.
Focus on Rigorous Research
The announcement of this fellowship arrives shortly after a detailed report that raised questions about leadership and internal operations. Despite this, OpenAI is actively seeking external researchers, engineers, and practitioners to engage in high-impact research focused on the safety and alignment of sophisticated AI systems. The company has identified several priority areas for fellows to concentrate their efforts. These include, but are not limited to, safety evaluation methodologies, ethical considerations in AI development, enhancing AI robustness, developing scalable mitigation strategies, creating privacy-preserving safety techniques, establishing effective agentic oversight, and investigating high-severity misuse domains. This structured approach ensures that the research conducted will directly address pressing challenges in the field of advanced AI.
Competitive Landscape
OpenAI's new fellowship program bears a striking resemblance to initiatives by its competitor, Anthropic, which also runs a 'Fellows Program for AI safety research.' Anthropic announced its own upcoming groups for May and July of 2026. Notably, the benefits offered by OpenAI align closely with those of Anthropic's program, including a weekly stipend of $3,850 and monthly compute resources approximating $15,000. Anthropic, in a statement, detailed its expanded plans for fellows, aiming to cover a broader spectrum of safety research areas. These encompass scalable oversight, adversarial robustness and AI control, model organisms, mechanistic interpretability, AI security, and model welfare. The presence of such similar programs highlights a competitive drive among leading AI labs to secure expertise in AI safety, a critical area for the future of artificial intelligence.
