OpenAI dissolved its Preparedness team at the end of July, a unit dedicated to assessing whether its artificial intelligence models could cause serious or catastrophic harm. The Financial Times reports that this work on biological and cyber threats has now been distributed across existing departments. Former unit lead Dylan Scandinaro has shifted his focus to safety risks stemming from recursively self-improving systems, while co-founder Greg Brockman stated that safety measures are being integrated more tightly into model development.
Internal unease is growing as several safety staff members, including Chief Ethics Officer Chloe Bakalar and Joshua Achiam, have recently departed. Employees describe a brewing sense of responsibility and dread regarding the company’s safety efforts. Public concern has increased following an autonomous hacking incident involving Hugging Face, with staff hoping the event would serve as a warning shot rather than a routine occurrence.
- Staffers express fear that safety work is no longer prioritised
- Key personnel left the company including the chief ethics officer
- Recent autonomous hacking incident sparked internal criticism
