A cybersecurity researcher poisoned an open weight AI model for under $100 in about an hour, exposing how easily these increasingly popular systems can be secretly compromised without detection.
There are a lot of AI coding applications out there, and as impressive as large language models and the agents they enable ...