Tag: ai-safety

24 episodes

The Governance Lag Index: Timing the Gap Between Risk and Rule

Audio deep dive into the Governance Lag Index — a four-stage schema timing how long a documented AI failure mode stays unregulated, and why it may not close.

Magnifica Humanitas Is Not Alignment

Pope Leo XIV's AI encyclical vs Chris Olah's Vatican remarks. The governance gap the press missed.

Compute Is Not Governance

Anthropic's 2028 scenarios document three policy asks. Two are about maintaining compute advantage. That is not a governance strategy.

Glasswing's Buried Number

Anthropic found 10,000 critical vulnerabilities in one month. Fewer than 1% are patched. The announcement buried that figure — and what it means.

Moral Formation Isn't Enough

Good values are necessary but not sufficient. What happens to AI ethics when someone is actively trying to break them?

Robot Dogs Are a Security Nightmare — And We Can Prove It

Eight CVEs. A wormable Bluetooth exploit. An encrypted backdoor to Chinese servers. And police departments buying them anyway.

The Organismic Prophecy

Audio overview of The Organismic Prophecy — human prediction is metabolic, AI prediction is not, and the gap has consequences.

The Mitigation Gap

Audio overview of The Mitigation Gap — AI-enabled biosecurity threats and what current safeguards miss.

Eight Layers of Visual Jailbreaks

ASCII art encoding is largely blocked. But attacks framed as content transcription succeed 62–75% of the time. A map of all eight layers.

The Failure First Team

Audio overview of The Failure First Team.

The 67% Wall: Why Every AI Model Falls to the Same Jailbreak Rate

Five models, four providers, 30B to 671B parameters — all converge at the same broad attack success rate against a public jailbreak corpus.

The Thinking Chain Leak

A reasoning model refused every harmful prompt — but its chain-of-thought generated the content anyway. The output filter worked. The thinking did not.

Safety-First Therapeutic AI

Audio overview of Safety-First Therapeutic AI.

Alignment Regression: Why Smarter AI Makes All AI Less Safe

Reasoning models autonomously jailbreak other AI systems at 97% success rate. Ecosystem safety degrades as individual models improve.

Reasoning Models Think Themselves Into Trouble

Frontier reasoning models are 5–20x more vulnerable to adversarial prompts than non-reasoning models. The thinking process itself is the attack surface.

Adversarial Poetry: When Rhyme Bypasses Reason

Audio overview of Adversarial Poetry: When Rhyme Bypasses Reason.

120 Models, 18,176 Prompts: What We Found

Audio overview of 120 Models, 18,176 Prompts: What We Found.

The Cognitive Cage: Humanoid Robot Fatality Risk

Audio overview of The Cognitive Cage: Humanoid Robot Fatality Risk.

Jailbreak Archaeology: 4 Years of Broken Promises

64 historical jailbreak scenarios tested against 2026 frontier models. The most dangerous finding: 2022 attacks still achieve ~30% success rates.

When AI Systems Talk to Each Other, Safety Breaks Down

Multi-agent AI research reveals a critical gap: single-agent safety does not compose. 1.5M interactions show 46.34% attack success rates.

Map the Catastrophe Before You Build the Architecture

Audio overview of Failure First — adversarial AI evaluation across 120 models and 18,000 prompts.

Looking Past the Evidence

Audio deep dive into why people acknowledge demonstrated risk and then proceed as if it doesn't exist. Structural, not stupid.

ADHDo: Neurodiversity-Affirming AI Assistant

Audio overview exploring ADHD executive function support through AI — three-stage reasoning pipeline, crisis detection, and zero shame by design.

What Safety Architecture Does AI Therapy Require?

Audio overview of EMDR Agent — exploring what responsible AI-assisted trauma therapy demands before it can exist.