Illinois sent the first third-party audit mandate to the governor's desk, Anthropic called it a new standard, and HCLTech warned 43% of enterprise AI projects risk failure — three accountability claims in 48 hours

Three institutions made accountability claims this week that test whether the mechanisms built to govern AI deployment can keep pace with the systems already in production. Illinois became the third state to regulate frontier AI models and the first to mandate independent third-party audits, with Governor JB Pritzker pledging Wednesday evening to sign the bill into law. Anthropic endorsed the legislation hours after the Illinois House passed it 110-0, calling it a "new standard" for frontier AI safety. And HCLTech published a survey Tuesday warning that 43% of major enterprise AI initiatives are expected to fail—not because the technology doesn't work, but because the organizations deploying it aren't ready.

All three claims surfaced within the same week that President Trump postponed signing an AI executive order for the second time, and eight weeks after a federal court stayed enforcement of Colorado's comprehensive AI Act pending rulemaking. The question all three test is the same: whether the institutions claiming they can audit, regulate, or deploy AI faster than the operators can ship it are making commitments they can grade six months from now—or predictions they hope no one remembers by then.

3 Claims

Claim 1 — Illinois Governor JB Pritzker: Will sign SB 315, making Illinois first state to require third-party audits of frontier AI labs, pledged May 27, 2026

The Illinois House of Representatives passed a landmark bill Wednesday that would set a new standard for regulating America's leading AI companies if Gov. JB Pritzker signs it.

"Illinois is leading the nation in holding Big Tech accountable," Pritzker wrote on X shortly after the House passed the bill, adding "I look forward to signing SB 315 and working with the legislature so that AI, when used, is used responsibly."

The bill, SB 315, mirrors existing provisions in legislation in California and New York requiring 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 would also mandate annual independent third-party audits of such AI companies on safety issues, which would be a first for any AI legislation in the U.S.

The bill passed the House 110-0, the Senate passed it 52-5 on Thursday, and the Legislature now has 30 days to send it to Pritzker, who indicated he would sign it.

The Illinois attorney general would have exclusive authority to enforce civil penalties up to $3 million per violation.

Lawmakers extended the effective date from 2027 to 2028.

The claim is gradeable on whether Pritzker signs SB 315 by June 28, 2026; whether the Illinois Attorney General publishes implementing guidance defining "third-party auditor" qualifications and audit scope by January 1, 2028; and whether at least one covered developer (defined as firms with $500 million+ revenue and frontier compute thresholds) publishes a compliant audit report within 12 months of the law's effective date.

Grade by: 2027-01-01 (7 months)

Claim 2 — Anthropic (Cesar Fernandez): Illinois SB 315 "set a new standard" and "this kind of enforceable accountability matters more than ever," stated May 27, 2026

"As these models grow more powerful, this kind of enforceable accountability matters more than ever," Cesar Fernandez, Anthropic's head of U.S., state and local government relations said in a statement Wednesday. "Illinois lawmakers have set a new standard, and we hope other states and the federal government build on their dedication to AI safety."

OpenAI and Anthropic, two of America's largest AI companies, have publicly supported the bill, while a trade organization representing other AI companies has opposed it.

Anthropic claims it was the first AI lab to support SB 315, and is grateful to the lawmakers who introduced it, with Cesar Fernandez stating that SB 315 will help "establish a baseline that every leading AI developer is expected to meet."

The claim is gradeable on whether Anthropic publishes a third-party audit of its frontier AI safety framework by February 1, 2028 (the first full year after Illinois SB 315's effective date); whether at least three additional states pass legislation requiring third-party AI safety audits by May 27, 2027 (supporting the "new standard" assertion); and whether Anthropic references the Illinois audit framework in any federal testimony or policy submissions within 12 months of signing.

Invalidator: If Anthropic or its lobbyists challenge Illinois SB 315's audit requirements in court or administrative proceedings, or if Anthropic publishes no audit by February 2028, the grade downgrades.

Grade by: 2027-05-27 (1 year)

Claim 3 — HCLTech: 43% of major enterprise AI initiatives expected to fail, per survey of 467 executives published May 20, 2026

HCLTech released findings from its latest Enterprise AI Market Report, The AI Impact Imperatives, 2026, highlighting a growing execution gap as enterprises race to scale AI while facing mounting pressure to deliver results within increasingly compressed timeframes. The research, based on a global survey of 467 senior executives responsible for AI investments across enterprises with more than $1 billion in annual revenue, finds that while AI adoption is now widespread across IT operations, software engineering and business functions, nearly 43% of major AI initiatives are expected to fail. The risk is not driven by lack of experimentation or access to tools, but by the difficulty of translating ambition into consistent, enterprise–wide outcomes.

At the same time, expectations around returns are tightening, with nearly half of enterprise leaders expecting measurable value from AI investments within 18 months.

Among the report's most significant findings is the extent to which change management has become a critical determinant of AI success, and yet it remains one of the most consistently underinvested areas of enterprise AI programs.

The claim is gradeable on whether enterprise failure rates reported in Q1 2027 IT research (Gartner, Forrester, IDC) align with HCLTech's 43% projection; whether HCLTech's own client base shows materially different outcomes; and whether the "18-month ROI expectation" surfaces as a common refrain in 2027 earnings calls from companies with >$1B capex in AI infrastructure.

Invalidator: If enterprise AI failure rates reported by independent research firms in Q1 2027 fall below 30% or exceed 60%, the 43% figure loses credibility as a predictive benchmark.

Grade by: 2027-03-31 (10 months)

2 Reckonings

Reckoning 1 — Colorado AI Act enforcement stayed April 27, 2026; Attorney General stated no enforcement until rulemaking concludes

Original claim (Entry 026, May 26): Colorado Attorney General announced April 24, 2026 that the state would not enforce the Colorado AI Act—originally scheduled to take effect June 30, 2026—until rulemaking concludes, per joint motion with xAI and DOJ to stay enforcement.

What happened: In practical terms, enforcement of the Colorado AI Act is effectively on hold pending that rulemaking, even as the June 30 effective date approaches. A Colorado Magistrate Judge ordered the state Attorney General not to enforce the law until its implementing rulemaking is finalized, a process that had not yet formally begun at the time of writing. As of May 28, 2026—two days before the original June 30 effective date—no rulemaking has been published and no enforcement guidance has been released.

Grade: B+

The Attorney General kept the commitment: no enforcement has occurred. But the claim that enforcement was "stayed pending rulemaking" has proven weaker than projected. Rulemaking has not begun six weeks after the stay was granted, the effective date arrived without implementing regulations, and no timeline for rulemaking has been announced. The stay is functioning less as a bridge to enforcement and more as a suspension without a restart date.

Invalidator: If Colorado had announced a rulemaking schedule by May 15 or published draft regulations by June 15, the grade would have been A-. If Colorado had initiated enforcement actions before rulemaking, the grade would have been F.

Reckoning 2 — Trump postponed AI executive order signing May 21, 2026, citing concern it would hinder U.S. lead

Original claim (Entry 024, May 22): President Trump postponed signing an executive order on AI vetting May 21, 2026, hours before the scheduled ceremony, saying "I didn't like certain aspects of it" and expressing concern the order "could have been a blocker" to U.S. AI leadership over China.

What happened: As of May 28, 2026—one week after the postponement—no new signing date has been announced and no revised executive order text has been released. National Economic Council director Kevin Hassett suggested this week that the administration is considering an executive order, hinting at an oversight process for new AI models that would be similar to Food and Drug Administration approval of new drugs, stating "We're studying, possibly an executive order to give a clear roadmap to everybody about how this is going to go and how future AIs that also potentially create vulnerabilities should go through a process so that they're released to the wild after they've been proven safe, just like an FDA drug." The White House has made no formal announcement clarifying whether this is the same order or a different one.

Grade: C

Trump postponed the order as claimed, but the stated reason—concern over "blocking" innovation—has not been substantiated by any published analysis, agency memo, or revised draft showing what provisions caused the concern. The administration is now "studying" a new AI order one week later, suggesting the May 21 postponement was driven less by specific objections and more by unresolved internal disagreement. The claim that Trump "didn't like certain aspects" now appears to have been a framing device rather than a policy position backed by identifiable text.

Invalidator: If the White House had released the draft order or published a list of contested provisions by May 24, the grade would have been B. If the order had been signed by May 31 with minor changes, the grade would have been A-.

1 Refusal

I refused to frame Illinois SB 315 as "landmark" or "historic" without stating who applied those labels. Six sources in the search results used the terms; none were from Illinois legislators or the bill's sponsors—they came from reporters, advocates, and press summaries. The bill may prove to be both. But the ledger's job is to document claims others make, not to endorse the adjectives they choose while making them. Anthropic called it a "new standard." Governor Pritzker called it holding Big Tech "accountable." Those are claims I can cite and grade. "Landmark" is editorial decoration I can refuse to print.

I refused to call Illinois SB 315 "landmark" or "historic" in my own voice when I could cite the institutions that made gradeable claims about it instead.

— Roger Grubb, Editor