Malicious actors and state-linked groups in China and Russia are already attempting to use Anthropic’s artificial intelligence for cyberattacks, mass surveillance, and other dangerous operations. This is stated in a new company report, Politico reports, writes UNN.
Details
Anthropic said it had detected and blocked a number of cases involving the misuse of Claude models. These included cyber operations, surveillance systems, influence campaigns, as well as attempts to use AI for research that could potentially be linked to the creation of biological weapons.
The company recorded cases in which its models were targeted for use in automating cyberattacks and intelligence operations. Some schemes were linked to government agencies or entities from China and Russia.
Biological Threat
Anthropic was particularly concerned by requests related to biology. In one case, a user attempted to obtain help from AI with research that could have made a dangerous virus more harmful. The company said it had blocked the activity.
Anthropic is also countering attempts to illegally copy Claude’s capabilities to train other models. The company previously reported on a large-scale illegal ecosystem whose participants were attempting to circumvent restrictions and create a large number of accounts to gain access to its technologies.
The company warns that as AI capabilities grow, such systems may increasingly simplify the execution of sophisticated cyberattacks and other dangerous operations that previously required a significant number of specialists and resources.