RAI.Brief· Byron Arnao
UPDATED 14 MIN AGO · THU 29 MAY
"
Safety is shifting from research to infrastructure. OpenAI's Rosalind launch and Anthropic's $65B raise mean the safety conversation now lives in product roadmaps and government contracts — not just papers.
KEY INSIGHT · MAY 29 · 7-DAY READ
Global Safety Index
68
/ 100
↑ 3 pts above 30-day average of 65
7d↗ +3 30d→ 0 YTD↗ +11
EU AI ActImplementation
↗ safer
US RegulationState patchwork
→ mixed
Safety ResearchLab investment
↗ safer
Cloud FrameworksAWS · Azure · GCP
↗ maturing
Top 5 · this week RANKED BY IMPACT × FRESHNESS
01
Research · OpenAI · May 29

OpenAI ships safety as a product with Rosalind Biodefense

The first major lab to take safety out of the paper and into a government-partnered deployment — a quiet inflection point for the industry.

GPT-Rosalind expands defensive access for biological threat preparedness, with Commerce Department coordination. The frame shifts: safety is no longer just research output, it's infrastructure that gets sold, audited, and depended on.

openai.com · 3 min read Open
02
Legislation · EU · May 7

EU AI Act Omnibus extends deadlines, bans nudifiers

High-risk standalone systems pushed to Dec 2027. Non-consensual generative imagery and CSAM prohibited from Dec 2026.

europa.eu · 2 min Open
03
Capital · Anthropic · May 28

Anthropic at $965B valuation

$65B round makes it the world's most valuable AI startup. Funds scaled interpretability work.

guardian.com Open
04
State · US · May

7 US states pass AI laws as federal stalls

CT, CO, GA, NY, IL active. Trump's preemption EO delayed. Patchwork persists.

congress.gov Open
05
Framework · AWS · May 13

AWS ships ISO/IEC 42001 implementation guide

8-dimension RAI playbook with financial services GRC focus.

aws.amazon.com Open
MAY 19
Research · Nature
DeepMind Co-Scientist published in Nature

Multi-agent Gemini system for iterative hypothesis generation. Hassabis warns of "species-level transition" — calls for international coordination.

MAY 28
Framework · OpenAI
OpenAI Frontier Governance Framework

Aligns with California Transparency in Frontier AI Act and EU AI Act Code of Practice. Covers cyber-offense, CBRN, manipulation, loss-of-control mitigations.

MAY 1
Legislation · Connecticut
Connecticut SB 5 — AI Responsibility Act passes

Regulates AI in employment decisions. Phased rollout Oct 2026 → Oct 2027. Awaiting Governor Lamont's signature.

MAY 14
Legislation · Colorado
Colorado SB 26-189 signed, replaces CAIA

Shifts from high-risk classification to disclosure model. AI use in employment and housing must be disclosed to affected individuals.

MAY 16
Framework · Google
Google 2026 Responsible AI Progress Report

Seven-pillar framework: research, policies, testing, mitigation, launch review, monitoring, governance forums. Updated Frontier Safety Framework central to Gemini 3 deployment.

APR 10
Framework · Microsoft
Azure CAF for AI Governance expanded

Cloud Adoption Framework integrates NIST AI RMF Govern–Map–Measure–Manage loop. Frontier Governance Framework added for advanced model capabilities.

MAY 21
Policy · US Federal
Trump delays federal AI cybersecurity EO

Internal admin and industry disagreement over federal oversight scope. Preemption strategy execution incomplete; states continue legislating.

MAY 25
Voice · Anthropic
Olah meets Pope Leo XIV on AI ethics

Anthropic co-founder participates in Magnifica humanitas encyclical event. "Questions AI raises extend beyond computer science."

Sources
Scanned
EU Commission · congress.gov · Anthropic · OpenAI · DeepMind · AWS · Microsoft · Google · Nate B. Jones · Latent Space · Hard Fork · All-In · Lenny's Podcast · Dwarkesh · Lex Fridman · Pivot · Nature
42briefings
17sources
11podcasts