Output is not evidence
Three pieces of AI security research landed in a single month, on three different problems. They all point at the same soft spot.
Topic
AI development security
AI coding assistants and agents in real engineering teams: permissions, review capacity, evidence and governance.
AI writes the code. Who reviews the risk?
AI ships code faster and repeats old vulnerability patterns with confidence, while NIS2 raises the bar on how software is built. Why security belongs in design and review.