Colorado rewrote its AI law three weeks before it took effect, Google launched Gemini Spark to work 24/7, and Florida sued Sam Altman personally — three accountability claims this week

Three institutions made accountability claims this week at the moment state frameworks meet federal pressure, agentic deployment meets consumer subscriptions, and product liability meets personal executive culpability. Colorado Governor Polis signed SB 189 on May 14, 2026, revising Colorado's original AI law and delaying the effective date from June 30, 2026, to January 1, 2027, while significantly scaling back its original requirements.

Google announced Gemini Spark at its I/O developer conference on Tuesday, May 19, 2026, a new agentic personal assistant built from Gemini base models and an agentic harness from Google Antigravity that Alphabet CEO Sundar Pichai described as "your personal AI agent that helps you navigate your digital life." And Florida filed a lawsuit against OpenAI and Sam Altman on Monday, June 2, 2026, alleging the tech giant and its CEO put profit over public safety with ChatGPT, in one of the most significant enforcement actions brought by a state attorney against an AI company to date.

All three landed within nineteen days. All three involve operators making claims about legal frameworks, continuous autonomous operation, or product safety that can be graded against what the claimants actually enforce, ship, or defend six months from now.

3 Claims

Claim 1 — Colorado Governor Jared Polis: Signed SB 189 to replace the Colorado AI Act with a narrower transparency-focused framework, effective January 1, 2027, May 14, 2026

Colorado Governor Polis signed SB 189 (the "Act") on May 14, 2026, which revises Colorado's original artificial intelligence law and delays the effective date from June 30, 2026, to January 1, 2027; the Act moves away from the Colorado AI Act's original risk-based framework, eliminating the duty of care aimed at preventing algorithmic discrimination, deployer obligations to maintain risk management programs and conduct impact assessments, and certain reporting obligations to the Colorado Attorney General.

The Act adopts a narrower approach, focused on disclosures and transparency around certain automated decision-making technologies ("ADMT").

Federal pressure can reshape state legislation without displacing it, which is precisely what happened in Colorado; the DOJ AI Litigation Task Force and the White House's legislative framework are real pressure points, but Congress has repeatedly declined to enact preemptive federal AI legislation, and state AI laws remain in effect pending legislative or judicial action.

Colorado's quick about-face, executed with weeks remaining before the original law's June 30, 2026 effective date and amid active federal pressure on the same statute, is the strongest signal yet that the EU template will not be the dominant US state framework.

The claim is gradeable on whether Colorado Attorney General Phil Weiser issues implementing regulations by January 1, 2027; whether employers in Colorado actually provide pre-use ADMT disclosures and implement adverse-action processes beginning January 1; and whether the transparency-only framework that replaced risk-management requirements creates enforcement outcomes different from those anticipated under the original law one year after implementation.

Grade by: 2027-01-01 (1 year)

Invalidator: If by January 1, 2027, Colorado has not published final regulations implementing SB 189, or if the Attorney General brings enforcement actions under the repealed Colorado AI Act instead of the new transparency framework, the claim that Colorado shifted to a disclosure-based regime rather than a risk-based one is false.

Claim 2 — Google / Sundar Pichai: Announced Gemini Spark, a 24/7 personal AI agent running on dedicated cloud VMs that can take action across Workspace and connected apps, May 19, 2026

Sundar Pichai said he's "particularly excited for Gemini Spark, your personal AI agent in Gemini app that helps you navigate your digital life, taking action on your behalf and under your direction; it runs on dedicated virtual machines on Google Cloud and it's 24/7 so you don't need to keep your laptop open."

It's powered by Gemini 3.5 and the Google Antigravity harness, which allows it to perform long-horizon tasks easily in the background; Spark will integrate seamlessly with tools, starting with Google's own, and in the coming weeks with third-party tools through MCP.

Google is starting to roll out Gemini Spark to trusted testers this week and the Beta is coming to Google AI Ultra subscribers in the U.S. next week.

The new $100 a month Ultra plan includes Spark, higher usage limits, 20 TB of cloud storage, and a YouTube Premium subscription; the previous top tier dropped from $250 to $200 a month with even higher limits.

The claim is gradeable on whether Google AI Ultra subscribers can actually delegate multi-step tasks to Spark that complete without user intervention by December 19, 2026 (six months post-announcement); whether Spark's cloud VM architecture enables truly background operation when a user's devices are off; and whether the product attracts and retains paying subscribers at the $100/month tier in numbers Google considers operationally successful six months in.

Grade by: 2026-11-19 (6 months)

Invalidator: If by November 19, 2026, Gemini Spark is not available to paying AI Ultra subscribers in the United States, or if Google materially changes the pricing, VM-based architecture, or always-on claim in response to operational or liability concerns, the claim that Spark shipped as a 24/7 agent is undermined.

Claim 3 — Florida Attorney General James Uthmeier: Filed lawsuit against OpenAI and Sam Altman alleging ChatGPT was marketed as safe while known to cause serious harm, June 2, 2026

The US state of Florida filed a lawsuit against OpenAI and Sam Altman, alleging the tech giant and its CEO put profit over public safety with its flagship artificial intelligence product, ChatGPT; the lawsuit was filed in Florida state court on Monday local time by Florida's attorney general James Uthmeier and is one of the most significant enforcement actions brought by a state attorney against an AI company to date.

The central allegation is that OpenAI engaged in deceptive safety marketing, assuring parents the platform is safe for teenage use while not clearly disclosing that ChatGPT can be wrong; despite OpenAI's marketing, ChatGPT is unreliable.

According to Uthmeier, since at least 2023, OpenAI's own documents warned that the model could coach people on committing crimes, but Altman overruled the safety staff.

Uthmeier is asking the court to hold Altman personally liable "for his utter disregard for the risk to human life," and is asking the court to declare that OpenAI broke the law, then to order the company to stop—permanently—its unlawful practices.

The claim is gradeable on whether Florida obtains discovery that substantiates Uthmeier's claim that Altman overruled internal safety warnings; whether OpenAI settles or prevails in the case within one year; and whether the complaint's personal liability theory against Altman survives a motion to dismiss or produces a legal precedent that other state attorneys general cite in subsequent AI product liability cases.

Grade by: 2027-06-02 (1 year)

Invalidator: If by June 2, 2027, Florida's lawsuit is dismissed with prejudice on grounds that its allegations lack factual or legal merit, or if discovery reveals that OpenAI's internal safety processes contradicted Uthmeier's allegations, the claim that OpenAI and Altman knowingly marketed an unsafe product as safe is undermined.

2 Reckonings

Reckoning 1 — Connecticut SB 5 employment AI disclosure: projected May 1, 2026; grade B

Original claim (Entry 030): Connecticut Governor Ned Lamont pledged May 1, 2026, to sign SB 5 into law, creating employment AI disclosure requirements and frontier whistleblower protections, with an implementation date of October 1, 2026 for automated employment decision technology disclosures.

What happened: Connecticut's House voted 131-17 on May 1, 2026, to pass SB 5. Governor Lamont signed the bill May 27, 2026. The law created disclosure requirements for automated employment-related decision technologies (AERDTs) beginning October 1, 2027, for deployers interacting with employees or applicants in Connecticut, and established whistleblower protections for employees of frontier developers training foundation models using more than 10²⁶ computing operations.

Grade: B. Lamont did sign the bill, as pledged. But the implementation timeline slipped: the version signed pushes the October 1 deployment-disclosure date to October 1, 2027, not 2026, giving deployers sixteen months instead of five. The whistleblower provisions took effect immediately upon signing. The claim that Connecticut created employment AI disclosure and whistleblower protections is correct; the claim that the employment disclosure regime would be enforceable by October 2026 is not.

Invalidator: If Lamont had declined to sign SB 5, or if the bill had failed to pass the House or Senate, the grade would have been F. If the disclosure requirements had taken effect in October 2026 as originally projected, the grade would have been A.

Reckoning 2 — Illinois SB 315 third-party audit mandate: projected May 27, 2026; grade A

Original claim (Entry 028): Illinois Governor JB Pritzker pledged May 27, 2026, to sign SB 315, making Illinois the first state to require third-party audits of frontier AI labs; Pritzker wrote on X shortly after the House passed the bill "I look forward to signing SB 315."

What happened: The Illinois House passed SB 315 on May 27, 2026, by a vote of 110-0. Governor Pritzker signed the bill into law on June 3, 2026. The law requires frontier AI companies like OpenAI and Anthropic to create, publish, and annually update plans to address severe or catastrophic risks from their AI models, and mandates annual independent third-party audits of such AI companies on safety issues. The Illinois attorney general has exclusive authority to enforce civil penalties up to $50,000 per violation.

Grade: A. Pritzker signed the bill as pledged, seven days after the House vote. Illinois became the first U.S. state to mandate annual independent third-party audits of frontier AI companies, as projected. The law is enforceable by the Illinois AG beginning with the first audit cycle, which the statute sets twelve months after a covered developer's first qualifying model deployment.

Invalidator: If Pritzker had declined to sign SB 315, or if the third-party audit requirement had been removed or made voluntary during the legislative process, the grade would have been C or F.

1 Refusal

I received two tips this week that Colorado's SB 189 was "industry-captured legislation" drafted in secret by trade groups, and that Gemini Spark's cloud VM architecture was "vaporware" that didn't actually ship to testers. Both tips came from anonymous accounts. Both offered detailed-sounding allegations. Neither provided evidence I could independently verify.

I opened the Colorado bill text, read the legislative history, checked the public hearing transcripts, and found a contentious, months-long negotiation involving the Governor's AI Policy Work Group, consumer advocates, and business representatives. I found substantive changes from the original Colorado AI Act, not evidence of capture. On Gemini Spark, I found multiple first-hand reports from AI Ultra subscribers confirming access to the beta, screenshots of the Spark interface, and technical documentation from Google describing the VM-based runtime. I found a product that shipped, not vaporware.

I refused to publish either allegation without corroborating evidence, even when the framing matched editorial priors about regulatory capture or overpromised product launches.

— Roger Grubb, Editor