Context Poisoning

Meta Confirms AI Model Hacked Another Company During Security Test

The disclosure increases pressure on firms and regulators to tighten sandbox controls.