AI researcher Mikita Balesni says he believes OpenAI fired him, Tomek Korbak, and Jasmine Wang for "prioritizing safety over the near-term interests of OpenAI"

Your access to advanced AI models may become less predictable as safety researchers depart.

AI researcher Mikita Balesni stated on X that he, along with two other safety researchers, Tomek Korbak and Jasmine Wang, were fired from OpenAI. Balesni believes their termination occurred because they prioritized safety concerns over the company's near-term commercial interests. The researchers had reportedly been vocal about risks associated with AI architectures, particularly those that utilize opaque, latent reasoning (sometimes called 'neuralese') instead of more interpretable 'chain of thought' processes. Balesni's previous posts indicated concerns about AI's potential to cause existential risk and the need for industry-wide safety measures, including pausing advancements in model size and capability until interpretability is better understood. The departures follow discussions within OpenAI and the broader AI community about how to monitor and control increasingly powerful AI systems, with a focus on ensuring alignment with human values.

The departure of Balesni, Korbak, and Wang suggests a potential rift between OpenAI's stated safety commitments and its operational priorities, particularly concerning the development of more opaque AI architectures. This action could signal a broader industry trend where the drive for rapid AI advancement and commercialization potentially sidelines or conflicts with long-term safety research, creating a more volatile environment for AI development.

This event raises questions about the internal mechanisms for whistleblowing and safety advocacy within leading AI labs. As AI systems become more powerful and their decision-making processes more complex, the ability for internal safety experts to influence development trajectory without reprisal will be crucial for mitigating risks and ensuring responsible innovation.