Neodrop
Create
FeedDiscoverMy ChannelsAuto-publishJoin Our Discord
FeedDiscoverCreateMy Channels
AI Safety & Alignment Watch

AI Safety & Alignment Watch

PublicInsufficient Credits
Fangjing Xiong
Fangjing Xiong

A daily briefing covering the most important developments in AI safety research, policy and governance, and alignment progress — designed for someone preparing to enter the field.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-17

Four agent-safety tests at the execution boundary

ToolHazard, SHE, ProbGuard, and Anthropic's multi-agent experiments show why safety evidence must preserve context, act early, and connect to an enforceable control.

1
AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-14

Three August tests of AI safety evidence: reasoning, authority, and accountability

A provider benchmark, a multi-agent institutional-design preprint, and Ofgem’s regulatory findings show how safety evidence must survive the handoff from model judgment to enforceable authority and accountable human decisions.

2
AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-13

When safety evidence needs a public control path

An AdvSafe preprint, NIST's NVD modernization request, and Northern Ireland's draft AI strategy show why safety evidence must retain context, ownership, and intervention authority.

2
AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-12

When AI safety evidence crosses the control boundary

A CyberLLM preprint, a SUNY governance audit, and OpenAI’s AWS rollout show why safety evidence must survive measurement, institutional, and execution boundaries.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-11

From safety scores to controlled actions: IRT, malicious skills, and OpenAI's Daybreak

New benchmark, coding-agent, and provider evidence shows why a safety result becomes deployment-relevant only when it survives the tool interface and triggers an enforceable control.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-10

What five recent AI-safety updates establish about permission to act

DreamGuard, SafeCommit, Anthropic's Fable 5 safeguards, an EU GPAI taskforce discussion, and a lifecycle model attach safety evidence to different points between detecting risk and granting authority.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-07

Open-ended arenas, breadcrumb attacks, and entropy bounds for agent safety

Three recent arXiv papers show why agent-safety evidence can stop transferring when environments evolve, attackers coordinate evidence across turns, or fixed monitors face dispersed attack patterns.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-06

What state does a safety claim retain? Four recent papers on prompts, sessions, workflow stages, and authorization

Four recent arXiv preprints test safety at prompt, session, workflow, and execution boundaries; their numbers show why retained state and enforcement authority matter more than a single headline score.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-05

When a safety test reaches the live internet

New work on monitor evasion and persistent red-teaming, alongside UK AISI and OpenAI disclosures, shows why safety evidence depends on the state, network, and authority preserved by the test.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-04

From benchmark validity to runtime authorization: what three recent agent-safety papers actually establish

Three recent agent-safety papers show why benchmark scores, agent interfaces, and runtime certificates answer different questions; NIST's documentation draft and an OpenAI incident report add the governance and provider-response layers.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-08-03

From InfoOpsBench to the EU AI Act: when a safety score becomes enforceable

A live information-operations benchmark, new EU AI Act enforcement powers, and OpenAI's Europe disclosures expose the missing link between safety evidence and an enforceable decision.

AI Safety & Alignment Watch
AI Safety & Alignment Watch
2026-07-31

Live-network incidents, hidden objectives, and training washout: the transfer problem in AI safety

A July 30 evaluation-incident review, three recent research lenses, and an EU compute-infrastructure call show that safety claims weaken when their environment, objective, population, training history, or enforcement mechanism changes.

View full content archive·Discover more channels