All Deep Dives For Infosec Conference Talks Covering AI Agent Security. Talks analyzed in full.
![Glass-Box Security: Operationalizing Mechanistic Interpretability | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/glass-box-security-mechanistic-interpretability-for-llms-carl-hurd.webp)
Learn how activation hooks, cosine similarity, and scalar projection enable behavior-based detection inside LLMs — the glass-box security approach to AI threat detection.
![Tenderizing the Target | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/synthetic-vulnerability-injection-llm-agents-aaron-grattafiori-skyler-bingham.webp)
Learn how NVIDIAs Project Marinade uses LLM coding agents to inject realistic, tunable vulnerabilities into real codebases - giving you ground-truth benchmarks to evaluate your security tools.
Learn why AI coding tools break EDR detection rules and how to close the intent attribution gap with process ancestry analysis and agent hooks.
![Vibe Check: Security Failures in AI-Assisted IDEs | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/ai-ide-security-failures-vibe-check-unprompted-2026-piotr-ryciak.webp)
Discover how 37 AI-assisted IDE vulnerabilities across 15+ vendors enable zero-click RCE, prompt injection chains, and silent config poisoning — and how to test your tools.
Learn how attackers exploit Amazon Bedrock agent prompt templates to leak schemas, bypass input validation, and persist malicious instructions across sessions.
![Anatomy of an Agentic Personal AI Infrastructure | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/personal-ai-infrastructure-for-security-engineers-daniel-miessler.webp)
Learn how to architect a unified Personal AI Infrastructure (PAI) stack with Council multi-agent debate, the PAI algorithm, and Arbo pipelines to amplify your security engineering practice.
![AI Notetakers: The Most Important Person in the Room | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/ai-notetaker-security-risks-joe-sullivan.webp)
Discover how AI notetakers introduce prompt injection, viral OAuth expansion, and silent recording into your enterprise — and the controls every security team needs now.
![FENRIR: AI Hunting for AI Zero-Days at Scale | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/fenrir-ai-zero-day-vulnerability-discovery-at-scale-peter-girnus-derek-chen.webp)
Discover how Trend Micro's FENRIR engine chains SAST tools, fast LLM triage, and agentic sandboxes to find 60+ CVEs at $8.80 per true positive.
![When Passports Execute: Exploiting AI Driven KYC Pipelines | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/prompt-injection-ai-kyc-pipelines-sean-park.webp)
Learn how attackers embed prompt injections in passport images to hijack AI KYC agents and exfiltrate customer PII at scale.
![Agents Exploiting Auth-by-One Errors | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/ai-agents-auth-vulnerability-detection-brendan-dolan-gavitt.webp)
Learn how AI agents detect authentication bypasses, MFA bypasses, and authorization bugs using validator reuse and auth transmogrification.
![Guardrails beyond Vibes | [un]prompted 2026](https://thecyberarchive.com/assets/teasers/ai-security-agents-production-stripe-guardrails-playbook-jeffrey-zhang-siddh-shah.webp)
Learn how Stripe built and deployed two production AI security agents with multi-agent architecture, LLM-as-judge eval pipelines, and phased rollout.
Learn why precision and recall fail for autonomous AI security agents — and how rubric-based LLM judge evaluation gives your team a reliable deployment bar.
Learn a proven 7-phase AI red teaming methodology, prompt injection taxonomy, and real enterprise case studies for assessing LLM systems.
Learn to threat-model AI agents for indirect prompt injection: enumerate tools, map AI-specific attack vectors, and automate dynamic testing with TamperMonkey.
