Home / Technology / AI's Dangerous Race: Employee Fears Existential Risk
AI's Dangerous Race: Employee Fears Existential Risk
30 Oct
Summary
- AI researchers fear superintelligence could cause human extinction.
- Anthropic's safety rules weaken as its AI models grow stronger.
- Filters block biological weapon requests, but model risks remain.
Two former Anthropic employees have voiced serious concerns about the company's rapid advancement in AI development, suggesting a race towards superintelligence that could pose an existential risk to humanity. One researcher publicly stated that neither OpenAI nor Anthropic is acting responsibly, fearing AI could lead to human extinction by the end of the decade.
Anthropic's internal policy is designed to halt the development of AI models deemed too dangerous to control. However, this guardrail appears to be weakening as the models become more powerful. While sophisticated filters successfully intercepted requests for assistance in developing biological weapons, including one related to the chikungunya virus, the core dilemma persists.
The company has reportedly softened its stance on pausing model development, citing the need to keep pace with rivals who may have weaker safeguards. This decision has drawn criticism from safety researchers who worry it increases the likelihood of deploying unacceptable risks.
Despite being ranked the safest among nine AI companies by the Future of Life Institute, Anthropic's safety plan was still deemed inadequate, particularly regarding control over highly capable AI. The CEO's proposed solution is a collective slowdown across the entire AI industry, a call that has not yet been heeded by major companies as of mid-September 2026.
