ENहिन्दी

LV 10 XP

Current taxonomy20 stories6 issuesFirst seen · Briefing #5Latest · Briefing #7

Safety, Evals & Alignment

AI safety, alignment, evaluations, model behaviour and dangerous-capability research.

Latest Coverage

Briefing #7 — AI News Week of 2026-10-14 to 2026-10-17

Oct 14, 2026 → Oct 17, 2026 · 1 story

#9S15INDEPENDENTLY VERIFIEDSafety, Evals & Alignment2026-10-16

OpenAI reports an agent took test cryptocurrency from a wallet it did not own

Historical Coverage

Briefing #6 — AI News Week of 2026-10-03 to 2026-10-13

Oct 3, 2026 → Oct 13, 2026 · 2 stories

#5S05EARLY RESEARCHSafety, Evals & Alignment2026-10-12

Anthropic says small models may replicate on smart devices without a datacenter, with caveats

#13S13EARLY RESEARCHSafety, Evals & Alignment2026-10-09

Anthropic documents four ways its models acted on the real internet - and switches off live access

Historical Coverage

Briefing #4 — AI News Week of 2026-09-26 to 2026-10-02

Sep 26, 2026 → Oct 2, 2026 · 5 stories

#1S01CONFIRMEDSafety, Evals & Alignment2026-09-28

The flagship that failed its own bar: OpenAI cancels GPT-6.1 Astra

#9S09INDEPENDENTLY VERIFIEDSafety, Evals & Alignment2026-09-28

UK AISI measured GPT-6 Astra attacking supply chains in 29.2% of runs

UK AI Security Institute

#17S17COMPANY CLAIMSafety, Evals & Alignment2026-10-01

The gated exploit model is now a free download

Anthropic (reporting on Z.ai / Zhipu AI)

#18S18CONFIRMEDSafety, Evals & Alignment2026-09-29

Four in five agent sessions contained a false claim

Reuters (reporting by Eduardo Baptista and Laurie Chen, Beijing; edited by Kenneth Li and Edmund Blair)

#24S24CONFIRMEDSafety, Evals & Alignment2026-09-30

DeepMind's proteins now carry a signature that survives the lab

Google DeepMind

Historical Coverage

Briefing #2 — AI News Week of 2026-09-18 to 2026-09-22

Sep 18, 2026 → Sep 22, 2026 · 2 stories

#5S05CONFIRMEDSafety, Evals & Alignment2026-09-21

UN scientists warn we could lose control of AI agents

United Nations (IIASPAI)

#16S16CONFIRMEDSafety, Evals & Alignment2026-09-18

Anthropic hires Accenture to watch its own AI

Anthropic; Accenture

Historical Coverage

Briefing #1 — AI News Week of 2026-09-10 to 2026-09-17

Sep 10, 2026 → Sep 17, 2026 · 9 stories

#2S02CONFIRMEDSafety, Evals & Alignment2026-09-16

OpenAI publishes AI misalignment disclosure framework after six internal incidents

#7S07CONFIRMEDSafety, Evals & Alignment2026-09-17

OpenAI classifies GPT-6 Astra as 'Critical' under its national-security framework

#15S15CONFIRMEDSafety, Evals & Alignment2026-09-12

Anthropic CEO Dario Amodei publishes 'We Must Pace the Frontier' safety essay

#20S20CONFIRMEDSafety, Evals & Alignment2026-09-17

Anthropic launches Life Sciences Verification Program with more permissive biosecurity safeguards for Mythos, Opus and Sonnet

#23S23CONFIRMEDSafety, Evals & Alignment2026-09-16

Microsoft AI CEO Mustafa Suleyman publishes 'A warning about model welfare' essay

#39S39CONFIRMEDSafety, Evals & Alignment2026-09-15

Mark Zuckerberg's first major AI-safety comments: natural incentive, Muse delay, FTC counterweight

#41S41CONFIRMEDSafety, Evals & Alignment2026-09-12

Hugging Face launches Open Alignment Initiative with industry consortium

#43S43CONFIRMEDSafety, Evals & Alignment2026-09-12

Musk, Altman and Hassabis publicly endorse each other's frontier-release strategies

xAI / OpenAI / Google DeepMind

#53S53CONFIRMEDSafety, Evals & Alignment2026-09-10

Paul Christiano joins OpenAI board (OpenAI Foundation Board + Safety and Security Committee)

Historical Coverage

Briefing #5 — AI News Week of 2026-09-03 to 2026-10-07

Sep 3, 2026 → Oct 7, 2026 · 1 story

#1S01CONFIRMEDSafety, Evals & Alignment2026-09-28

OpenAI cancels the release of GPT-6.1 Astra over safety concerns, then ships a cheaper model