What the layoff record actually shows about AI in Washington

Tech's share of layoffs grew fast

Group the WARN filings by employer and the shift is hard to miss. In 2024, employers you would call tech accounted for about 3% of all reported layoff workers in Washington. In 2026 so far they account for about 44%.

The names are the ones you would expect. Since the start of 2025, Amazon has filed 7 notices covering 5,289 workers, Microsoft 6 covering 3,807, and Meta 4 covering 1,995. Google filed once for 52 workers, Zillow once for 91, Expedia once for 162.

• Washington ESD WARN requirements

A share can grow because the other part shrank

Percentages move for two reasons, and it matters which one is happening. Tech's 2026 share is high partly because tech filings went up, and partly because 2025 was an unusually heavy year for agriculture. Farms, orchards and packing houses accounted for 42% of reported workers in 2025 and only 16% so far in 2026.

Look at the raw counts instead and the picture is calmer. Tech employers reported 6,305 workers in 2025 and 7,039 in 2026 to date. That is an increase, but not the cliff the share number suggests on its own.

• Check the official ESD WARN record

What a WARN notice does not say

This is the part worth being careful about. A WARN notice records that an employer announced a large layoff. It does not record why. No filing in this archive says the word AI, and the form does not ask.

So when a headline attributes a specific layoff to automation, that reasoning comes from the company's own statements or a reporter's sourcing, not from the public filing. Both can be right. They are just a different kind of evidence than the record itself.

The filings also only cover larger employers meeting specific thresholds, so smaller layoffs never appear at all, and an announcement is not proof that every listed job ended.

• How WARN notices work

What researchers are finding, and how confident they are

The most-cited work here is a Stanford Digital Economy Lab paper by Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, which tracks payroll records for tens of millions of US workers. Their August 2026 update reports that employment for workers aged 22 to 25 in the most AI-exposed occupations sits roughly 19% below where it would be had it tracked similarly aged workers in less exposed jobs.

That finding is specific, and so are its limits. The same researchers say they do not see widespread economy-wide displacement, and they name the problem directly: it is hard to separate AI from interest rates, pandemic over-hiring and ordinary business cycles.

A Stanford policy brief from the same institution puts the other side plainly. Unemployment among the most AI-exposed workers rose 0.77 points since 2022 while the least exposed rose 0.85, which looks more like a soft labor market than a targeted one. Firms that adopted enterprise AI saw employment grow 10% over the following two years.

Both things can be true. Entry-level hiring can weaken in exposed fields while total employment holds up. That is a harder story to headline, and probably the more accurate one.

• Washington labor market information

How to read the next headline

When you see a number about AI and jobs, three questions get you most of the way. Is this a count or a share, and if it is a share, what happened to the denominator? Does the source establish why, or only what? And is the comparison group people in the same job market, or the whole economy?

The WARN record is good at what. It is silent on why, and it is worth saying so rather than filling the gap.

• Search the official ESD WARN database