A weekly executive briefing for AI executives, technology leaders, and heads of AI, covering regulation, policy, standards, and the AI governance vendor market.
- No. 14
When AI crosses a boundary, does your governance hold?
This week’s Governance Signal tracks AI risk at the transition points: first EU AI Act enforcement actions targeting containment evidence, a new harmonized standard that raises the bar beyond ISO 42001, China’s immediate export-registration duty, state scrutiny of AI data in bankruptcy, and Utah’s unresolved liability question for clinical pilots. Six developments, six executive takeaways, and a dated action list for your CISO, trade counsel, quality lead, procurement, and general counsel.
[Read the issue - No. 13
AI Accountability Requires Evidence
Can your organization explain an AI decision and prove its controls work? The October 13 Governance Signal examines proposed developer liability, human oversight, global regulatory developments, and vendor assurance. Includes five practical actions for AI leaders in a four-minute read.
[Read the issue - No. 12
AI Policies Are Not Proof of Control
Verifiable safeguards are becoming an executive priority. This issue covers the White House’s voluntary AI accord, California’s worker protections, OpenAI’s EU text-watermarking rollout, Utah’s independent healthcare AI evaluations, and accountability demands from New York City and the UN. Includes five practical actions for leaders.
[Read the issue - No. 11
Loss of Control Hits the Agenda
Agent containment is now a board-level issue. Covers New York's 72-hour incident-reporting mandate, the UN panel's warning on rogue agents, and a nine-week countdown to the EU's content-marking deadline. Includes five steps to act on now.
[Read the issue - No. 10
The States Take the Lead
With federal AI legislation stalled, governors are setting the rules. Covers California's kill-switch executive order, Virginia's new AI Task Force and data-center framework, and China's draft agent-safety standard. Shows why your compliance plan should be built around Sacramento and Richmond, not Washington.
[Read the issue - No. 09
Enforcement Gets Real
Regulators on three continents stopped asking for policies and started asking for proof. California signed its first AI audit laws, the EU AI Office sent its first formal information requests to 30+ model providers, and China issued AI liability rules. Includes five actions to build an evidence binder before a deadline forces one.
[Read the issue - No. 08
Twenty-Four Bills, One Signature Each: The California Countdown Begins
24 California AI bills. 30 days. One governor. If SB 574 gets signed, your legal team's AI workflow needs verification and disclosure controls before January 1. The UK just told AI vendors: not this bill, the next one. Agent governance is now an FY27 line item.
[Read the issue - No. 07
Sacramento Closed. Newsom Holds the Pen, January 1 Is the Deadline
California just made Sacramento the most important AI-policy venue on earth. Newsom has 30 days. OpenAI set a 30-minute monitoring bar. Anthropic's sandbox leaked into the real world. If you don't have a signing tracker and a network-isolation attestation, start both today.
[Read the issue - No. 06
Two Labs, One Week, Opposite Answers: Governance Just Hit the P&L
OpenAI paused training. Anthropic said it didn't have to. Same week, same threshold, opposite calls. One of them just added 20% to your unit cost. Your vendor questionnaire is a month out of date.
[Read the issue - No. 05
California Shipped Ten Bills. The Federal Ceiling Is Classified
Ten California AI bills advanced. One died. OpenAI just made Article 5 your problem. The federal frontier framework is complete and classified. Build to Illinois, the highest bar, and everything else becomes configuration.
[Read the issue - No. 04
The First Enforcement Data Points Are In
Enforcement has data now. The FLOP gate is presumed, not self-certified. Four kids' AI bills cleared committee. California's 30-bill suspense vote hits August 13. Agent governance just crossed $255M in three weeks. If your agent inventory isn't current, the market has already priced your gap.
[Read the issue - No. 03
The Clock Hit Zero on Both Sides of the Atlantic
Enforcement is live. EU Article 50 and California SB 942 switched on the same day. Preemption slipped to September, so state law is your US regime through Q3. If you can't produce an agent inventory with a kill-switch path, you're a design cycle behind.
[Read the issue - No. 02
The AI Office Is Now a Real Regulator
Six days out. The AI Office now has inspections and fines. Article 50 guidance says your AI agents must disclose themselves. An autonomous AI hack just rewired the Senate's August agenda. If your plan hasn't changed this week, it's already stale.
[Read the issue - No. 01
The EU Clock Hits Zero
August 2 is 14 days out. The EU pushed high-risk deadlines to 2027, but transparency rules and 7%-of-turnover fines still land on schedule. If your plan doesn't cover Article 50, it's incomplete. Five moves to make this week.
[Read the issue
Get each edition as it publishes.
Follow on LinkedIn for the weekly edition and commentary on AI governance, risk, and board readiness.
Also from Nancy Yaklich: The Ventured AI Brief, a biweekly brief for leaders turning enterprise AI into measurable business value.