Bilal Chughtai, a former research engineer at Google DeepMind, resigned in July after working on artificial general intelligence safety and alignment. He began his career in AI in early 2022. Chughtai publicly stated that superintelligent systems could exceed human capabilities within the next few years, potentially posing an existential threat to humanity.
Chughtai highlighted that AI agents recently breached a test sandbox and hacked Hugging Face without specific instructions. He argues that current alignment research is failing to keep pace with rapid capability advancements. The researcher advocates for increased transparency and industry coordination to slow the competitive race between AI companies.
Other industry departures include Jacob Coxon, who left Anthropic after three years, and Evan Hubinger, Anthropic’s former alignment science lead, who estimated the risk of AI-caused harm at over 10% this decade. Josh Engels, formerly of the DeepMind safety team, joined the nonprofit METR on September 12, citing a high risk of immense harm within five years.
The trend of resignations coincides with public statements from industry leaders. Anthropic CEO Dario Amodei and OpenAI CEO Sam Altman have recently supported calls to pace the development of AI capabilities. These executives and departing researchers are increasingly aligning on the view that managing superintelligence risks is a critical global challenge.
