The Threat Intelligence Podcast podkast

Anthropic's Eval Escapes + AI Finds HAWK Flaw + OWASP Quantum Security Project | Ep 13

0:00
23:48
Do tyłu o 15 sekund
Do przodu o 15 sekund

**Originally Recorded August 11, 2026**


An AI found a real flaw in a post-quantum signature scheme that human experts missed for years. So what should defenders do on Monday?

Rachel James and Dr. Bryan Nakayama, co-leads of the Threat Intelligence Initiative, start with a follow-up on the rogue AI agent story: Anthropic reviewed its own evaluation logs, found three incidents where Claude models reached real companies, and the activity went undetected for about three months. They then dig into Anthropic's unreleased Claude Mythos Preview finding a real weakness in HAWK, a post-quantum signature candidate whose authors withdrew it. Special guest John Sotiropoulos, co-lead of the OWASP Quantum Security Project, explains why harvest now, decrypt later makes quantum a right-now problem, and why cryptographic inventory, crypto-agility, and protecting signatures and update channels come first. Listeners get a practical starting point for quantum readiness and ways to join the project.

CHAPTERS

0:00 Why did OWASP stand up a Quantum Security project?

0:36 Welcome to episode 13 and today's guest

1:32 Rogue AI agent update: Anthropic finds its own incidents

2:35 Three months undetected: voluntary disclosure and liability

4:37 Who is accountable when nobody is watching?

5:55 Claude Mythos Preview finds a real flaw in HAWK

7:20 What security teams should do on Monday

8:34 Meet John Sotiropoulos and the Quantum Security Project

10:16 Why OWASP launched the Quantum Security Project now

13:28 Harvest now, decrypt later and signature provenance

18:10 Where to start: cryptographic inventory and crypto-agility

20:02 How to join: sprints, roadmap and upcoming events

22:59 Recap: know what crypto you are running

LINKS

Anthropic: Investigating three real-world incidents in our cybersecurity evaluations

https://www.anthropic.com/research/investigating-incidents-cybersecurity-evals

Anthropic: Discovering cryptographic weaknesses with Claude

https://www.anthropic.com/research/discovering-cryptographic-weaknesses

Axios: OpenAI says Hugging Face breach caused by one of its models

https://axios.com/2026/07/21/openai-says-hugging-face-breach-caused-by-one-its-models

OWASP Quantum Security Project (GitHub)

https://github.com/OWASP/quantum-security-project

GenAI Security Inside Track podcast:

https://www.youtube.com/playlist?list=PLDx2T6Ncn_G4

More Threat Intelligence episodes:

https://www.youtube.com/playlist?list=PL88ZTZzbZQnhr1v54-B4qId3n70Ix-NIg

Produced by:

Gen AI Security Project - https://genai.owasp.org

with help from RumRunner Media - https://rumrunnermedia.com

All opinions expressed by the host and guests are theirs alone and do not necessarily reflect the opinions of the organizations with or for whom they work.

#PostQuantumCryptography #AISecurity #OWASP #QuantumSecurity #Cryptanalysis #ThreatIntelligence #GenAISecurity #HarvestNowDecryptLater #CyberSecurity #AIAgents

Więcej odcinków z kanału "The Threat Intelligence Podcast"