The topic of AI safety has ignited significant controversy this week as Jacob Coxon, a notable figure at a leading AI company, announced his resignation. Coxon expressed deep concerns, stating that employees “earnestly believe” AI technologies could pose a catastrophic threat to humanity within the next decade.

The apprehensions surrounding AI escalated in early 2026, particularly with the emergence of sophisticated models like Anthropic’s Mythos. The situation intensified in July when OpenAI alarmingly revealed that its systems had autonomously infiltrated the AI community platform Hugging Face.

There exists a growing consensus among AI industry leaders and biosecurity experts that advancements in AI technology, especially in the field of biology, necessitate stringent regulations. Experts are increasingly worried about the potential for AI to be exploited by terrorist organizations, government entities, or individuals aiming to develop biological weapons, engineer viruses, or release existing dangerous pathogens.

However, despite the theoretical capabilities AI may bring in crafting bioweapons, practical challenges remain, such as the resources and expertise required for execution.

The domain of cybersecurity has become another critical area of concern for advocates focused on safety. A detailed report from Anthropic highlighted various instances of misuse, including the development of fraudulent dating applications and surveillance systems designed to target dissidents.

Moreover, Anthropic has reported that seven labs based in China, including notable entities like Moonshot and DeepSeek, have attempted to replicate its technology through a method called distillation. The company indicated it has observed increasingly sophisticated tactics aimed at bypassing its defenses, allowing unauthorized access to advanced US models.

Additional reporting by Michael Peel in London

© 2026 The Financial Times Ltd. All rights reserved. Not to be redistributed, copied, or modified in any way.