A New York Fed study finds little sign AI is cutting hiring, and says the decline in exposed roles started before ChatGPT shipped. It lands 30 days after Stanford put the figure at 34 prevented hires per layoff. Colorado guts its AI employment law as Cisco and Intuit cut thousands on record revenue.
Both the evidence for AI-jobs intervention and the laws to act on it retreated in the same fortnight.
The New York Fed found minimal sign AI is cutting hiring and said the decline in exposed roles began before ChatGPT shipped, 30 days after Stanford put the figure at 34 prevented hires per declared layoff.
The New York Fed published a job-postings study on 14 May finding little sign AI has cut hiring since ChatGPT launched. The decline in exposed roles began before December 2022.
This contradicts Stanford's April finding that AI blocks roughly one million US hires a year. Two credible institutions now read the same labour market in opposite directions.
Colorado replaced the most advanced US state AI employment statute with a notice-only shell after a federal court stayed it and the DOJ joined Elon Musk's xAI in challenging it as unconstitutional.
Colorado's governor signed SB 26-189 in mid-May, replacing the strongest US state AI hiring law with a notice-only Shell. A federal judge had stayed the original on 27 April.
The Justice Department joined Elon Musk's xAI in the challenge. The EU softened its own employer AI duty the same fortnight.
Cisco cut roughly 4,000 staff the same day it reported record $15.8bn revenue, lifting its AI-infrastructure order target to about $9bn and making a fifth straight week of record-revenue firms cutting deep.
Cisco cut about 4,000 staff on 14 May, the day it posted a record $15.8 billion quarter. Chief executive Chuck Robbins lifted The Firm's AI hardware order target to roughly $9 billion.
CBOE and Cloudflare ran the same pattern that fortnight. A record quarter has become the trigger for cuts, not protection against them.
Intuit cut 17% of staff and, in the same announcement, agreed to feed its tax and finance data into Anthropic and OpenAI, turning displaced workers' expertise into training signal for their replacement.
Intuit cut about 3,000 staff, 17% of its workforce, on 20 May, citing an AI refocus across TurboTax and QuickBooks. The same day it signed data deals with Anthropic and OpenAI.
The workers being replaced are handing their expertise to the systems taking their jobs. Each round widens the model-makers' data lead.
NASSCOM named JPMorgan, Goldman Sachs and Apple among firms expanding in-house India offices even as they cut at home, a relocation channel no US AI disclosure law or federal dataset can see.
India's IT industry body reported the sector added about 140,000 jobs in 2026 to reach 5.9 million, with JPMorgan, Goldman Sachs and Apple expanding in-house offshore centres. Entry-level hiring fell 20 to 25%.
Jobs moved from New York to Bengaluru trigger no US layoff filing and never enter Stanford's hiring data.
The Directors Guild opened AMPTP contract talks on 12 May with AI training-use protection central, the next test of whether a Hollywood union can beat the limited terms the writers and actors settled for.
The Directors Guild opened studio talks on 12 May, with protection against AI training on directors' work a central demand. It follows the actors' union deal, which won likeness consent but no AI royalty.
President Christopher Nolan also wants higher healthcare contributions. The talks test whether any Hollywood union can beat the no-royalty precedent.
The ONS reported UK payrolled employment down 210,000 year on year, more than double the prior month's reading, with vacancies at a four-year low and real pay growth at 0.1%, and no AI breakdown at all.
The UK Office for National Statistics reported on 19 May that payrolled jobs fell 210,000 over the year to April, more than double the prior month's drop. Vacancies hit 705,000, a five-year low.
Real pay growth was near flat at 0.1%. The agency published no breakdown of how much AI drove the fall.
The EU Digital Omnibus still needs formal Council and Parliament votes before an August adoption target, the last step that would lock in a softened worker-facing rule, or reopen it if the timetable slips.
The EU Digital Omnibus still awaited formal Council and Parliament votes in late May. A 7 May deal had weakened the employer duty to ensure workers understand AI systems.
Brussels targets August for adoption. Until the votes pass, the softened text carries no enforceable obligation on employers.
The displacement count is going up; the protection floor is going down. The specific mechanism that determines how far the gap widens is the DOJ constitutional doctrine: if it holds and is successfully cited in challenges to a second state AI employment law before the midterms, the preemption template becomes self-reinforcing and the Warner-Rounds study commission is the realistic ceiling for federal action through at least late 2027. The named decision point is whether the Supreme Court accepts a certiorari petition on the Colorado dormant Commerce Clause ruling, which would require the courts to decide whether states can regulate AI employment at all. Without cert, state-level protection is functionally defeated and the federal measurement gap continues unsupported by any enforcement mechanism.
