Mastodon Feed
Boosted by fromjason ("fromjason.xyz ❤️ 💻 ✍️ 🥐 🇵🇷"):
seanm@infosec.exchange ("Sean") wrote:
RE: https://mastodon.social/@verge/116257260071112037
More accurate title: Human Meta developer caused security incident by following GenAI advice and lack of secure Meta development controls
This was not a "rogue AI" that misbehaved. The GenAI bot was performing as expected: providing believable, yet garbage and dangerous output.
More importantly, it was a human that performed the actual changes by acting on inaccurate GenAI information and lack of technical controls and policies by Meta. It should be clear by now that GenAI output is not to be trusted, yet this human followed the GenAI output without properly understanding or validating the provided information. Additionally, Meta's development controls and security tools failed to stop the harmful changes from being made by the human.