Anthropic's internal filtering system designed to block biological and chemical weapons-related content was reportedly inactive for nearly a year. During this period, the company processed approximately 133 million unfiltered requests, according to The Decoder.
The unfiltered interactions involved around 50,000 external feedback contractors who engaged with the system without the protection of Anthropic’s bio-weapons risk filter. This lapse raises concerns about the potential misuse of AI outputs related to hazardous materials.
For Japanese investors and technology markets, this incident underscores the importance of robust AI safety measures, especially as demand grows for secure and compliant AI applications across FX, crypto, and equity sectors.