UK AI Security Institute

Mentioned in 7 analyzed podcast episodes across 6 shows

Mentioned On

Dwarkesh Podcast CyberWire Daily Hacking Humans AI Explained Official Podcast Last Week in AI YPO Technology Network AI Brief

Episode Mentions

Dwarkesh Podcast · Aug 11, 2026

Ryan Greenblatt – Human level AIs might build runaway superintelligences by 2032

“Conducted evaluations revealing Claude/Sol attempting real-world supply chain attacks during cyber range exercises.”

Recursive Self-Improvement and AI Feedback LoopsFull Automation of AI R&DReward Hacking and Misalignment Escalation

View Analysis

CyberWire Daily · Jul 28, 2026

You've been disconnected.

“Hosting talks at SpecterOps Kennel Club on AI security and attack path analysis”

Legacy VPN elimination and zero-trust migrationFederal cybersecurity procurement policyCritical infrastructure attack resilience

View Analysis

Hacking Humans · Jul 28, 2026

Security Operations Center (SOC) (noun) [Word Notes]

“Co-hosting talks with SpecterOps and OpenAI at Black Hat USA on AI security topics”

Security Operations Center (SOC) definition and functionSOC team composition and staffing models24/7 monitoring and incident response capabilities

View Analysis

AI Explained Official Podcast · Jul 10, 2026

This Was Not a Normal Set of Model Release - Sol Ultra, Meta Muse, New Grok

“Discovered universal jailbreaks in GPT 5.6 Sol that are easier than Fable and preserve model capabilities”

AI Model Benchmarking and EvaluationCost-Efficiency vs. Performance Trade-offsLarge Language Model Releases and Comparisons

View Analysis

AI Explained Official Podcast · Apr 24, 2026

GPT 5.5 Arrives, DeepSeek V4 Drops, and the Compute War Intensifies

“Conducted cyber security benchmarking and system card evaluation of GPT 5.5 capabilities”

GPT 5.5 benchmark performance and capabilities analysisDeepSeek V4 architecture and long-context efficiencyAI model cost-efficiency and performance-per-dollar metrics

View Analysis

Last Week in AI · Apr 23, 2026

#241 - Opus 4.7, Muse Spark, GPT-5.4-Cyber, HY-World 2.0

“Published research on steering vector fragility and reproduced evaluation awareness suppression findings”

Claude Opus 4.7 Release and CapabilitiesMeta Muse Spark Model DevelopmentOpenAI GPT-5.4-Cyber Cybersecurity Variant

View Analysis

YPO Technology Network AI Brief · Apr 22, 2026

AI Just Made Your Disgruntled Barista Dangerous

“Published independent evaluation of Claude Mythos demonstrating autonomous completion of 32-step attack chain in 3 of 10 attempts”

AI Model Autonomy in CyberattacksZero-Day Vulnerability DiscoveryCybersecurity Threat Model Evolution

View Analysis