Anthropic reports attempts to bypass safeguards for bioweapons research
Anthropic said it had stopped multiple attempts this year to use its technology for research that could assist in developing biological weapons. The company described five cases in which actors allegedly circumvented its controls or obscured the purpose of their work to avoid triggering safeguards.
The report said some of the users were in countries that Anthropic bars from accessing its models, including Russia, China and Iran. It did not identify the users or provide further details about the research in the supplied account. The cases come as experts express growing concern about the possible public-safety risks associated with artificial intelligence. Anthropic said biological work that may be dangerous can resemble legitimate scientific research, making it difficult for AI systems and their safeguards to distinguish harmful activity. The company said it published the examples to encourage discussion among AI companies and governments about emerging biological risks and ways to address them.
Written by Sakhanda Wire AI from reporting by Ars Technica. Facts, quotes and full context are in the original article below.