Topic
This topic groups publications by subject, regardless of whether they originate from professional work or public participation. One publication may belong to more than one topic.
Generative AI makes plausible claims, code and submissions cheap to produce. The deeper problem is what happens when verification still consumes scarce human time.
A practical evidence standard for AI-assisted vulnerability reports: scope, reproduction, PoC, demonstrated impact, severity, data minimisation and accountability.
Generative AI accelerates recon, hypotheses and reporting, while scope judgment, validation, impact analysis and accountable disclosure remain human responsibilities.
How to interpret AI benchmark results without overclaiming: test scope, data provenance, language coverage, uncertainty, reproducibility and intended-use evidence.