You might check this: https://news.ycombinator.com/item?id=44268335 https://news.ycombinator.com/item?id=44268335 Though note the following from the article l

You might check this: https://news.ycombinator.com/item?id=44268335 https://news.ycombinator.com/item?id=44268335 Though note the following from the article linked: "As long as both agents and their defenses rely on the current class of language models, we believe it is unlikely that general-purpose agents can provide meaningful and reliable safety guarantees."