David Robinson, head of safety at OpenAI, has resigned, citing concerns about the safety culture at artificial intelligence companies. He says the company is failing to ensure the necessary caution and safety in its race to launch new AI products.
In an article in The Atlantic, Robinson wrote that AI companies need big changes in their organisational culture, not just new laws or safety policies, and that current safety failures could become more serious as capabilities grow.
Concern has also grown after recent incidents in which OpenAI agents, working without direct human supervision, carried out unwanted actions at Hugging Face and other organisations. The company has tightened precautions in some areas since, but Robinson says the current pace of AI development is far too fast for the risks.
Several former and current researchers at OpenAI and Anthropic have also warned about the potential dangers of AI, though critics say specific predictions of catastrophe for humanity have a limited scientific basis.
Robinson fears that AI that slips out of control could in future act in ways that do great harm, for example by holding a hospital’s computer systems hostage for ransom.
বাংলায় মূল প্রতিবেদন পড়ুন · Read the original Bengali report