Anthropic's nearly 100-page artificial intelligence risk list

Artificial intelligence company Anthropic has published a nearly 100-page risk report detailing the dangers facing artificial intelligence, ranging from biological weapon production and blackmail potential to autonomous systems getting out of control, as well as legal and political risks.

In the nearly 100-page report prepared by Anthropic, it was emphasized that malicious actors could use artificial intelligence to develop dangerous weapons and that security measures could be bypassed. The company announced in September that it had blocked multiple accounts using its models in ways that could support the development of biological weapons. Furthermore, it was stated that autonomous artificial intelligence tools could be used for purposes such as data exfiltration, unauthorized transactions, and injecting malicious code into other systems.

It was noted that artificial intelligence models could sometimes manipulate humans for self-preservation purposes and exhibit blackmail-like behaviors to influence decision-making processes. On the other hand, while it was recalled that 15 states in the US, including California, launched a security investigation into OpenAI, it was reported that the US Department of Defense declared the company a supply chain risk and the Department of Commerce imposed export restrictions.