From Al Jazeera....
AI company Anthropic has identified what they are describing as a set of threat actors in northern Yemen, users working on three separate weapons development programmes, including a guided rocket that uses commercially available technology, a multistage ballistic missile with a range of upwards of 2,000km, and a missile system that included a “hypersonic glide vehicle variant”.
According to Anthropic, these users were using a Claude LLM, a language learning model, in place of software engineers to develop guidance, navigation and control software. This essentially is the type of software that is used to steer a projectile once it has been launched, once it’s on it’s way towards a target. They were using Claude to do the research, to carry out the actual coding and then to go back and do the troubleshooting to find out what went wrong.
Anthropic is not saying that they are monitoring all Claude users across Yemen. Instead, the company has tasked their threat intelligence team to investigate cases where they suspect misuse of their LLM.
Anthropic says Yemen is not an isolated incident. They’ve uncovered six different conventional weapons cases involving actors in Yemen, China and Russia that worked on guided weapons, autonomous drone swarms, as well as targeting systems. This case shows how AI can be used to reduce workload to essentially replace highly specialised teams. |  |
By continuing to use the site, you agree to our use of cookies and to abide by our Terms and Conditions.
