The paper behind this CAMLIS talk studies 24 model-and-runner configurations on a production TypeScript codebase whose reported vulnerabilities were non-public at scan time. DeepSec first screens files, then investigates candidates and revalidates reports; retained findings undergo root-cause mapping, independent source review and isolated proof confirmation. A separate four-model experiment compares minimal prompts, engineered prompts and structured source context across two runs. The important qualification is that the headline gain combines prompt engineering with CodeQL context. Relative to the engineered prompt alone, the primary CodeQL context improved yield only for the smallest tested model and reduced it for the other three. Models also found different subsets of the observed findings. The study's single main repository, small repeat count and differing runners limit generalization. Its reported scan costs exclude shared independent validation, and coverage refers to discovered root causes rather than every vulnerability actually present.
Vulnerability discovery: compare each model with minimal, engineered and enriched prompts
Related research
More curated notes connected through AI Red Teaming and Model Evaluation.
Pacing model development in an era of cyber-critical capabilities
OpenAI says preliminary evidence that Astra may meet its Critical cybersecurity threshold led it to pause frontier reinforcement-learning work for two weeks and keep its largest planned run on hold. New safeguards include stronger workload and network isolation, continuous boundary testing, token-level monitoring that escalates suspicious tool activity, and broader alignment checks for deception, reward hacking, and unauthorized access.
OWASP Top 10 for Agentic Applications for 2026
OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.
A blueprint for democratic governance of frontier AI
OpenAI proposes a three-part U.S. frontier-AI governance model: harmonize emerging state safety laws into a federal baseline, strengthen CAISI as an evaluation and standards institution, and coordinate a broader resilience program. Proposed controls include severe-risk evaluations, transparency reports, independent audits, safety-incident reporting, model-weight security, whistleblower protection, and periodic technical assessments.