When vulnerability discovery speeds up, response needs a queue, not panic
A practical playbook for turning faster findings into verified exposure, safer patches, and tested detections.
2026-11-09 · 5 min · 5 sources
AI SAFETY · AGENT SECURITY · GOVERNANCE
The people, ideas and evidence shaping safe AI.
We track what the field’s credible voices are saying on AI safety, security and governance — then add the context, comparison and practical implications a repost never will.
The spectrum — nine editorial pillars
Latest analysis
All articles →AI R&D is accelerating. The control evidence is not keeping pace
2026-11-09 · 5 min · 5 sources
Agent memory needs a maintenance loop, not just more storage
2026-11-09 · 5 min · 3 sources
OWASP's 2026 LLM list shifts security from prompts to blast radius
2026-11-02 · 5 min · 3 sources
Anthropic is using models as both workers and watchers
2026-11-02 · 4 min · 3 sources
Nothing in this pillar yet. See all analysis →
Formats
Never a single-source summary.
Five ways to add what the originals did not: comparison, context, a fair map of disagreement, or a useful decision.
01
The Synthesis
Several voices touched the same topic this week. Connect them, show where they agree and diverge, and say what it means.
2+ sources02
The Explainer
One idea from a recent talk/episode is important but hard to follow. Explain it from first principles, then relate it to other work.
1+ sources03
The Practitioner Playbook
Turn what experts said about a risk (e.g. prompt injection, agent permissions) into concrete controls a team can apply.
2+ sources04
The Debate Map
Voices disagree on something (e.g. open weights, timelines, regulation). Map each position fairly, the cruxes, and what evidence would change minds.
3+ sources05
Voice Profile
Profile one directory entry: their core ideas, how their thinking has evolved, and their best recent work. Feeds the directory.
2+ sourcesFrom the directory
All 134 voices →Robert Miles AI SafetyCoreyoutubeClear technical primers and interviewsAlignment & interpretabilityDoom DebatesCoreyoutubeCompare competing argumentsFrontier risk & controlGovernance & standardsAI ExplainedCoreyoutubeFast orientation to major developmentsForecasting & synthesisThe Inside ViewCoreyoutubeExpert interviews and research contextFrontier risk & controlGovernance & standardsFAR.AI ResearchCoreyoutubeTechnical talks and workshop recordingsEvals & red teamingAlignment & interpretabilityCenter for AI SafetyCoreyoutubePrimary-source talks and announcementsFrontier risk & controlEvals & red teaming
How we work
Our editorial method →01 — Discover
Start from the directory
Each piece starts with recent videos, podcast episodes, blog articles and posts from the tracked voices.
02 — Synthesize
Add what’s missing
We trace underlying evidence, compare claims and turn useful findings into decisions readers can apply.
03 — Verify
Check before publishing
Quotes are checked against source text, kept short and linked. Failed drafts are held for review.