Keynote: Not a Forecast: AI-Enabled Cyber, 12 Months On

Value of the Information & Strength of the Argument

The talk provides valuable, first-hand insights from a leading AI company’s threat intelligence team, with concrete case studies and quantitative data. The argumentation is coherent, tracing the evolution of AI misuse from simple to sophisticated, and supports the claim that AI is now a core component of cyber attacks. The speaker acknowledges limitations, such as the rough nature of the AI enablement score, and encourages transparency, which adds credibility.

Scientific Rigor, Source Quality, Title Accuracy

The speaker relies on internal Anthropic data, which is not publicly available, but the methodology is described. The talk references MITRE ATT&CK, a well-known framework, and mentions a forthcoming report. The title accurately reflects the content, and the talk is presented at a professional conference (SANS AI Cybersecurity Summit). No external sources are cited, but the speaker’s position and the use of internal data lend authority.

The title accurately reflects the content: a retrospective on AI-enabled cyber threats over the past year, emphasizing that it is not a forecast but an observation of current trends.

8/10

The speaker is the Head of Threat Intelligence at Anthropic, providing first-hand data from monitoring banned actors. The talk includes specific case studies with dates and statistics, but relies on internal, non-public data and lacks peer review.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and overview of the talk's purpose.
• 02:00Case study 1: UK-based ransomware actor using Claude in March 2025.
• 05:00Case study 2: Russian-speaking cybercriminal using Claude Code for data extortion in May 2025.
• 08:00Case study 3: Chinese state-sponsored group using Claude for autonomous cyber intrusion in September 2025.
• 12:00Landscape analysis: 800+ banned actors, MITRE ATT&CK coverage, and AI enablement score.
• 15:00Discussion on the dual-use nature of AI for defenders and the changing risk landscape.
• 18:00Conclusion: three key takeaways and call for transparency from AI labs.

• Anthropic Threat Intelligence Report (forthcoming) — Mentioned as a forthcoming report with more details on the MITRE ATT&CK mapping and AI enablement score.

• MITRE ATT&CK — The framework referenced for mapping threat actor techniques.