OpenAI disbands team working on AI safety
What's the story
OpenAI has disbanded its Mission Alignment team, which was dedicated to ensuring the safety, trustworthiness, and alignment of its artificial intelligence (AI) systems with human values. The team was formed in September 2024 as part of OpenAI's internal unit working on AI alignment. The news was first reported by Platformer and later confirmed by TechCrunch.
Team's purpose
Team focused on AI systems' human intent
The Mission Alignment team was part of a broader industry effort to ensure that AI systems act in accordance with human interests. An OpenAI job posting for the team described its focus on "developing methodologies that enable AI to robustly follow human intent across a wide range of scenarios, including those that are adversarial or high-stakes." The team's work was critical in ensuring these systems consistently followed human intent in complex real-world scenarios and adversarial conditions.
Leadership change
Achiam appointed as OpenAI's Chief Futurist
Josh Achiam, the former head of OpenAI's Mission Alignment team, has been appointed as the company's Chief Futurist. In a blog post announcing his new role, Achiam said his goal is to support OpenAI's mission of ensuring that artificial general intelligence benefits all of humanity. He will be working with Jason Pruet, a physicist from OpenAI's technical staff.
Internal restructuring
Remaining team members reassigned to other roles
A spokesperson for OpenAI confirmed that the remaining members of the Mission Alignment team, a group of six or seven people, have been reassigned to different parts of the company. The spokesperson said these members are now engaged in similar work in their new roles. However, it is still unclear if Achiam will form a new team as part of his futurist role.
Company statement
Routine reorganization, says OpenAI
A representative from OpenAI said the disbanding of the Mission Alignment team was a part of routine reorganizations that occur in the fast-moving companies. The company had previously formed a "superalignment team" in 2023 to study long-term existential threats posed by AI, but that group was also disbanded in 2024.