Yemeni Cell Used Anthropic’s Claude ‘In Place of Human Software Engineers’ To Develop Missile Guidance Systems
https://www.ibtimes.co.uk/anthropic-claude-ai-yemen-weapons-development-1819256
Yemeni Cell Used Anthropic’s Claude ‘In Place of Human Software Engineers’ To Develop Missile Guidance Systems
https://www.ibtimes.co.uk/anthropic-claude-ai-yemen-weapons-development-1819256
2 Comments
Anthropic says the guided rocket test apparently failed, but that a weapons-development cell was already using Claude like a small team of software engineers. Different instances were reportedly assigned coding, research and review tasks, lowering the amount of specialised human expertise needed. As these models become more capable, the bigger question is how much they could accelerate weapons development for groups that previously lacked the engineers, tools or technical knowledge to attempt projects like this.
Kinda scary but in a few years they will have access to a BETTER abliterated model, without any safety restrictions, locally. A dark future where bad actors (like the houthis) misuse ai.
Is there even a way to prevent this from happening? We can’t stop ai development because china won’t, and as we go further, open models will get better and more accessible.
Only thing I can think of for the US being able to stop this is making the hardware to run it (including consumer hw) completely inaccessible to anyone and obliterating everyone else’s means of producing such hardware.