On September 23, Oregon Governor Tina Kotek signed Executive Order 26-26, Establishing Responsible Artificial Intelligence Procurement Standards for State Government. The policy is blunt: Oregon will buy, contract for, and use frontier Artificial Intelligence (AI) models that have received independent, third-party safety review, consistent with state and federal law.

The order tells State Chief Information Officer Terrence Woods to define what counts as frontier for Oregon procurement, set criteria for adequate third-party review, and return an implementation proposal within 90 days. It also tells the state to assess whether a kill-switch requirement for frontier models is viable. The order took effect immediately. Kotek will reassess it every three months. (Governor's Office — Oregon; GovTech)

Oregon is not inventing the lane alone. Officials said the state will look to peers as it writes the definition and the reviewer bar, including California's independent-assessment statutes and New York's Responsible AI Safety and Education Act registration clock, which starts pulling frontier developers onto a state roster in November.

Monday's file is still Salem's buy-side gate, not Sacramento's November recommendations remake. The useful desk question is whether your vendors can already show an independent safety review that would survive Oregon's 90-day criteria draft. (DailySynapse; GovTech)

Also this week, Gambit Security published an interim report on a live, financially motivated campaign that used three open-source AI harnesses (Strix for vulnerability search, Cairn for autonomous exploitation, Hermes for orchestration) against hundreds of online retailers. Between September 10 and 15 alone, 105 attack projects launched and at least 27 companies were compromised to some degree. More than 600,000 credit card records were taken from two victims. Marginal model cost sat around twenty-five dollars a company.

That is ops color beside Monday's procurement file, not a remake of Saturday's evaluation-probe map. Boards still need a resilience answer when cleanup routines wipe tables after extraction. (Gambit; SecurityWeek)

In Rome, Legislative Decree 160 of 9 September 2026 is now on the books and takes effect September 30. It tightens civil procedure around AI-linked damage: courts can order disclosure of technical documentation, logs, and risk-management records, and it creates a rebuttable presumption of causation when EU AI Act duties were breached.

Useful liability-desk color next to Oregon's buy gate. If a state or contractor starts requiring third-party safety review, the paper trail that decree forces into evidence is the same pile vendors should already keep. (D.Lgs. 160/2026; Il Sole 24 Ore)

Sunday already put Supply Chain Risk on the desk, now sent on Buttondown: the D.C. Circuit's 2–1 holding that good-faith safety refusals can still be a supply-chain risk under 41 U.S.C. § 4713. Saturday's Agents That Probe stays sent. Friday's Virginia Executive Order 22 stays sent: Spanberger's Data Center Accountability Framework and rapid-response AI Task Force. California's Executive Order N-9-26 stays: independent verification and a verified frontier kill-switch study, with recommendations due by November 16. The Cybersecurity and Infrastructure Security Agency (CISA) and National Institute of Standards and Technology (NIST) Interagency Report (IR) 8587 token guide stays put.

Those are the procurement statute, the forensic agent map, Richmond, Sacramento, and the federal token track. Monday's file is Oregon's frontier buy order. (Supply Chain Risk; Agents That Probe; Virginia; Governor's Office — California; CISA)

Thursday's Agencia Española de Protección de Datos (AEPD) agent breach notice, Wednesday's Korea Internet & Security Agency (KISA) guide, Tuesday's Altman monitorability wall and multi-lab pact, and Monday's Amodei evaluator desks stay on the desk.

They already had their mornings. Madrid, Seoul, the pact track, and permanent reviewers next to today's Oregon buy file. (AEPD; SEDaily; TechTimes; Amodei)

Before the next state, municipal, or critical-infrastructure contract that names a frontier model, put three blanks on one page. Can the vendor show an independent third-party safety review that would clear Oregon's coming criteria. Who owns the gap if the model is frontier under Salem's definition and the review is still vendor-written. What happens when a kill-switch viability ask lands next to an "all lawful uses" demand.

Write the questionnaire now. Put the same blanks next to every lab and integrator selling into public-sector work west of the Rockies, then copy it east. (Governor's Office — Oregon; GovTech)

The Move

Oregon's EO 26-26 says the state will not treat frontier AI as a normal software buy. Independent third-party safety review becomes the gate, with a 90-day clock for the State CIO to turn that into criteria.

Send procurement and the chief information security officer (CISO) a one-pager. Ask which current vendors could clear an independent review today, who owns the definition fight if "frontier" lands broader than expected, and how a kill-switch viability ask would sit next to existing usage policies.

If the answer is "we only sell into commercial," the blank still goes next to every contractor who might pull that model into Oregon work, or into any state copying Salem's gate.

Seven bullets. One move.