> daily_signal(2026_05_30)

A Utah county attorney took the vote on a 40,000-acre data center away from residents. A Senate coalition told the Pentagon that ad networks sold the location of deployed troops to adversaries. Oregon told data centers to pay 100 percent of their grid.

PickBits Daily Signal · Saturday, May 30, 2026

By Mark Pickering · 12 min read · May 30, 2026

// tl;dr

The week's news was a collision between communities and regulators trying to install rules around the AI buildout and the surveillance, financialization, and access-to-justice consequences of the buildout that is already running. A Utah county attorney closed the public-vote path on a 40,000-acre data-center campus by reframing the underlying approval as administrative. A Senate coalition told the Pentagon that the same commercial advertising-data networks that target you with shoe ads sold the location of deployed U.S. troops to adversaries. Oregon became the first state to make data centers pay 100 percent of the grid they need. The Justice Department charged a Google security engineer with using internal search-trend data to win $1.2 million on a public prediction market. The UK Home Office is deploying AI facial age-estimation on asylum seekers at Dover before Parliament has debated it. And Australia's labor tribunal said its caseload has grown 70 percent in three years as workers use ChatGPT to file claims they could not previously afford to bring.

A Utah attorney closed the referendum path on a forty-thousand-acre data center. A Senate letter told the Pentagon adtech sold troop locations to adversaries. Oregon made data centers pay 100 percent of their grid. The DOJ charged a Google engineer for trading on internal search data. The UK Home Office pointed facial AI at asylum seekers. And Australian workers are filing unfair-dismissal claims at a rate no tribunal was built to absorb.

1. A Utah county attorney denied residents the right to vote on a 40,000-acre data center. He ruled the commissioners' approval was administrative, not legislative, and not referable.

Continuing the data-center community-pushback arc that ran 5.20-5.28 (Maine, Denver, Minneapolis, Delaware-Fenwick LLC voting, Minnesota Skyway, Texas Hill County). Today the arc gets a new mechanism: the county attorney as the gate that closes the public-vote path before residents reach the ballot.

Box Elder County Attorney Stephen R. Hadfield rejected three citizen referendum applications this week, denying residents the right to put the Stratos Project data-center campus to a public vote. The campus is planned for 40,000 acres of unincorporated land in northern Box Elder County, south of US-84, and is backed by the Military Installation Development Authority (MIDA) together with private developers including former "Shark Tank" investor Kevin O'Leary. The three referendums sought to overturn commissioners' votes approving two resolutions: an interlocal agreement with the developers, and an associated zoning-and-services framework. Hadfield's ruling, issued Thursday, held that none of the resolutions were legally referable under Utah's Election Code, because the commissioners' actions were "administrative" acts implementing existing state statutes, not the creation of broadly applicable new laws.

The newly formed Box Elder Accountability Referendum group has signaled it will pursue a legal challenge to the attorney's ruling. The applicants have a 10-day window to file an appeal under Utah law, and the group's representatives told Utah News Dispatch that an appeal is being prepared. The ruling closes the direct-democracy path that residents in Saline Township, Pulaski County, Denver, and Minneapolis have used in the past month to fight data-center builds in their backyards, by reframing the predicate as an administrative not a legislative act. If the Box Elder ruling stands, the playbook for the next forty-thousand-acre data center is to structure the county approval as administrative implementation of state statute and route around any referendum threat from the start. The Saline Township legal playbook that ran on May 10 worked because the underlying act was legislative; the Box Elder playbook works because it is not.

Why this matters: If you live anywhere a data center has been proposed in your county or town, the legal mechanism that the people in Box Elder, Utah used to try to put it to a public vote was just denied on a reading that almost every other county in the country can copy. The county attorney's reasoning is the new variable: if the commissioners' vote can be reframed as "administrative implementation" of state law, the public referendum path closes before a signature can be gathered. Action this week: If your county is weighing a data-center project, ask your county attorney in writing whether the approval will be framed as administrative or legislative, because the answer determines whether the public has a referendum path at all; the question alone often flushes the answer into the record. If you are on a county or town council, the Box Elder ruling is the new model your county attorney will be tempted to copy; read the ruling and ask whether your jurisdiction is willing to defend the residents' referendum right. If you are a resident, write your county commissioner and ask them to publish, in advance of any data-center vote, whether they intend to take the action as administrative or legislative, and what that means for citizen review.

kuer.org: Box Elder County rejects data center referendums, but opponents aren't giving up
utahnewsdispatch.com: Box Elder County residents won't get to vote on data center referendum, county attorney says

KUER coverage of Box Elder County rejecting citizen referendums against the 40,000-acre Stratos data center
kuer.org · May 28, 2026

2. A Senate coalition told the Pentagon this week that advertising networks sold the location data of deployed U.S. troops to adversaries. The data came from apps on soldiers' personal phones.

On May 28, Senators led by Ron Wyden (D-OR) and a bipartisan group sent a letter to Pentagon leadership calling commercial data brokers and advertising networks "a national security threat." The letter followed a disclosure Wyden's office obtained: a letter dated April 14, 2026, from U.S. Central Command to the Senate Commerce Committee confirming that adversaries had obtained commercially available location data to identify and track U.S. military personnel in active deployment zones.

The mechanism is not a classified system breach. The data originates with ordinary smartphone applications: fitness trackers, navigation apps, weather tools, and free social platforms that collect precise GPS coordinates and sell them to advertising data brokers. Those brokers resell the data in bulk to commercial buyers with few restrictions on who the buyers are or what they do with the information. CENTCOM's April 14 letter confirmed that some of those buyers used the data to determine the physical locations of U.S. service members in the field.

The Wyden coalition's letter asks the Pentagon to assess which service members may have been affected, to restrict personal devices in certain operational contexts, and to support legislation capping the commercial sale of location data to foreign nationals. No legislation currently restricts what a data broker can sell to a foreign buyer operating through a domestic intermediary. The commercial location-data market that surfaces this information is the same one that sells advertising audiences to brands, political campaigns, and employers.

Why this matters: If you have ever installed a free weather app, a free navigation tool, or any app that asked for your location and said it used the data for "analytics" or "personalized experience," your location history has almost certainly been sold to at least one data broker. The CENTCOM case is about soldiers, but the mechanism is the same one operating on every phone in your household. The difference between a service member's location data and yours is the potential consequences of its misuse, not whether the data exists or who can buy it. Action this week: Open your phone's settings, find Location Services or App Permissions, and review which apps have "always on" location access. On iPhone: Settings > Privacy & Security > Location Services. On Android: Settings > Location > App Permissions. Revoke always-on access for any app that does not need it to function. Then look up your state's data-broker opt-out registry, which currently exists in California, Virginia, Colorado, and Texas, and submit your household's opt-out request.

militarytimes.com: US troops are reportedly being targeted using location data, Pentagon says

Military Times coverage of CENTCOM disclosure that adversaries used commercial location data to track deployed US troops, following Wyden-led bipartisan Senate letter to the Pentagon
militarytimes.com · May 28, 2026

3. Oregon became the first US state to make data centers pay 100 percent of the cost of expanding the electric grid they need. Sixteen data centers are already in scope.

Continuing 5.29 #3 (Pennsylvania GRID standards put the cost-allocation on developers via the tax-incentive lever). Oregon's path is the regulatory lever: a public-utility-commission tariff on the utility, not a deal on permitting. Same destination, different machinery.

The Oregon Public Utility Commission approved Portland General Electric's large-load tariff framework this week, formally signing off on the order the PUC first published on May 5. The tariff is the first regulatory framework in the United States to require data centers and other large-load customers drawing 20 megawatts or more to enter dedicated service contracts that put them on the hook for 100 percent of the distribution-network expansion costs their loads cause. Contracts also require them to use at least 90 percent of their contracted power capacity, and impose financial penalties for exceeding the contracted load or exiting the contract early. Projects above 100 megawatts pay an additional 1-cent-per-kilowatt-hour surcharge dedicated to programs that offset residential customer costs and address low-income energy burden. PGE has until June 3 to file the implementation pricing; the tariff takes effect June 10. Sixteen Oregon data centers are already in scope.

The decision ends an arrangement that for years had Oregon residential and small-business customers effectively subsidizing the technology sector's grid build-out. PGE rates are up nearly 50 percent over the last five years, with the utility spending roughly $210 million in 2025 on grid expansion in Hillsboro to serve data-center growth. Oregon's mechanism is regulatory, not legislative: the Pennsylvania GRID Standards published Wednesday route through tax incentives and fast-track permitting (a developer who wants the benefit accepts the cost-allocation rule); Oregon's path routes through the public utility commission's tariff authority and binds every customer above the load threshold by operation of law, regardless of whether the developer wanted state benefits. Both states arrive at the same outcome: data-center developers pay for the grid they need. The next US state to write a large-load tariff will read the Oregon order line by line.

Why this matters: If you live in Oregon and pay an electric bill, the additional generation that the sixteen data centers in your state need was being added to the bill you already pay, and that arrangement just ended. If you live in any other state, the regulatory template is now public and can be cited in your own state's public-utility-commission docket the next time a hyperscaler files for an interconnection. The wall-power asymmetry that has been pricing into your electric bill for two years just got its first regulatory answer in the United States. Action this week: Look up your state public utility commission's open dockets and search for "large load" or the name of any data-center developer with a project in your county; if a tariff filing is open, the Oregon order is now the comparable case to cite in a public comment. If you are at a utility, the Oregon order is the regulatory framework your interconnection team has been waiting for to refuse to socialize hyperscaler-driven expansion costs. If you are a resident, write your state representative this week and ask whether your state's public utility commission is opening a similar docket on large-load tariffs.

utilitydive.com: Oregon PUC approves PGE's large-load tariff framework for data centers

Oregon PUC approves PGE large-load tariff requiring data centers to pay 100 percent of grid expansion costs
utilitydive.com · May 29, 2026

PickBits Daily Signal is the working brief I file every morning at 6 a.m. ET. If a friend forwarded this to you and you want it in your inbox, you can subscribe at pickbitsai.substack.com. It is free.

4. The Justice Department charged a Google security engineer with using internal Google Search data to win $1.2 million on Polymarket. He bet against Pope Leo XIV.

The US Attorney for the Southern District of New York charged Michele Spagnuolo, a 36-year-old Italian citizen living in Switzerland and a Google information-security engineer of more than twelve years, with commodities fraud, wire fraud, and money laundering on Wednesday. According to the complaint, Spagnuolo used internal access to Google Search trend data, specifically the confidential telemetry that feeds Google's annual Year in Search marketing campaign, to place bets on the public Polymarket crypto-prediction market under the username AlphaRaccoon. The 2025 most-Googled-person market was the target. Prosecutors say Spagnuolo risked more than $2.7 million across positions and won approximately $1.2 million.

The specific positions tell the story. Spagnuolo allegedly bet nearly $1 million that Kanye West's wife Bianca Censori would not be the most-Googled person of 2025, more than $600,000 that Pope Leo XIV would not take the top spot, and a substantial sum on the rapper D4vd, who has been charged with murder, at a time when other Polymarket traders had priced D4vd at near-zero. Each of those bets, prosecutors argue, was made with confidential Google search-volume data the public did not have. Spagnuolo was arrested Wednesday, appeared in federal court, and was released on a $2.2 million bond with travel restrictions. The case is the second Polymarket insider-trading filing in six weeks. Polymarket has rapidly become a venue where information asymmetries from inside major tech companies, banks, and political organizations are alleged to be cashing out.

Why this matters: Every search query you type into Google is logged, aggregated, and sits behind the company firewall. This week the Justice Department alleged that an employee with access to that aggregate took it home and bet on a public market about your behavior, and won a million-plus dollars. The misuse is the news; the bigger story is that the legal scaffolding to call this commodities fraud was already there. If you wondered whether the things you Google were ever priced and sold for personal profit, the answer this week is yes, at least once. Action this week: Pull up myactivity.google.com, sign in, and turn off Web and App Activity if you have not already; the default-on setting is what builds the per-account telemetry an engineer with internal access could correlate. If you use Polymarket or any crypto-prediction market, the second insider filing in six weeks is the signal that the platform's information-asymmetry problem is now visible to federal prosecutors, which changes the regulatory horizon. If you work in security or compliance at any large tech company, the case is the template a US attorney will use the next time an employee correlates internal telemetry with a public market; the access logs and the bet records are the entire trail.

techcrunch.com: Google engineer charged with insider trading after making $1.2M on Polymarket

DOJ charges Google engineer Michele Spagnuolo with using internal search data to win $1.2 million on Polymarket
techcrunch.com · May 27, 2026

5. The UK Home Office is paying to scan asylum seekers' faces with AI to estimate their age. The contract went to a German facial recognition firm via a UK intermediary. Human Rights Watch says scrap it.

The UK Home Office awarded a £322,000 contract to Akhter Computers Ltd to pilot AI facial age-estimation on asylum seekers arriving at Dover. The underlying facial recognition technology comes from Cognitec, a German firm that Akhter is subcontracting. The system would scan the faces of individuals claiming to be children and return an age estimate before formal documentation review. The Home Office stated that in the year ending March 2026, more than 6,400 migrants claiming to be children were age-assessed at the border, with 43 percent found to be adults. Human Rights Watch called the approach "deeply flawed" and asked the government to scrap the contract.

The accuracy gap in facial AI on non-European faces is documented across multiple EU government procurement audits and is specific to Cognitec's systems: the error rate rises significantly for darker skin tones and non-Western facial structures, the demographic most represented in the asylum seeker population arriving at Dover. The error has a specific asymmetry: misclassifying an adult as a child adds a person to the protected minor immigration track; misclassifying a child as an adult strips that child of protected status. The pilot has not yet begun and Parliament has not held a debate on the contract.

Why this matters: If you have ever traveled internationally, been stopped at an immigration checkpoint, or helped anyone navigate an asylum or visa process, this is about AI making preliminary legal determinations about who receives child protection status. The Cognitec accuracy gap on non-European faces is not a theoretical concern. It is documented in EU government procurement records. The UK Home Office is deploying it on the most vulnerable population at the border before Parliament has reviewed the contract. Action this week: Read the Biometric Update coverage linked below. It includes the Cognitec subcontract detail that the primary LBC report omits, and the documented accuracy gap figures from prior EU deployments. If you work in immigration law, HR compliance for international employees, or civil liberties advocacy, use the standard UK Freedom of Information Act request template to ask the Home Office for the accuracy benchmarks cited in the procurement justification. If you work in enterprise procurement and evaluate biometric vendors, add the Cognitec accuracy-gap documentation to your evaluation criteria before the next contract review.

lbc.co.uk: AI facial recognition to check age of asylum seekers from next year

UK Home Office awards £322,000 contract for AI facial age-estimation on asylum seekers at Dover, subcontracted to German firm Cognitec
lbc.co.uk · May 29, 2026

6. Australia's labor tribunal said AI-generated unfair-dismissal claims have grown its caseload by 70 percent in three years. Workers who could not afford lawyers are filing on their own now.

The week's constructive-use story. AI lowering the cost of bringing a legal claim is helping a class of workers who were previously priced out of representation. It is also breaking a tribunal that was not built for the volume.

Justice Adam Hatcher, president of Australia's Fair Work Commission, told the Senate Estimates committee this week that the tribunal's caseload has jumped from roughly 30,000 matters in 2023 to about 45,000 in 2024-25, and is now on pace for between 50,000 and 55,000 lodgments in the current 2025-26 financial year. That is more than 70 percent caseload growth in three years. The Commission's own statutory 12-week processing target for unfair-dismissal claims has slipped. Hatcher attributed the growth primarily to generative AI: workers who could not previously afford a lawyer are now using ChatGPT and similar tools to draft and file claims, and lodging at a rate the tribunal has never had to handle.

The story has two sides. On the upside, the access-to-justice gap that has historically priced low-wage Australian workers out of an unfair-dismissal claim has narrowed sharply: the marginal cost of filing went from a paralegal-hour to a free chatbot session, and the number of claims being filed reflects the resulting demand. On the downside, a non-trivial share of the AI-drafted filings come in with hallucinated legal precedents, paraphrased statutes the AI made up, or filings outside the qualifying period the worker never knew about. The Commission says it will soon require every applicant to disclose whether they used AI to draft their submission and whether they checked the AI's work, and is lobbying the federal government for legislative powers to dismiss matters that are not properly prosecuted or that have no reasonable prospect of success. Australia is the first major employment tribunal to publicly document this trade-off, which is the framing the UK, EU, and US labor courts will read in the next twelve months as their own caseloads start showing the same curve.

Why this matters: If you have ever been fired and felt that the cost of bringing a claim was the reason you walked away, the Australian numbers say that for tens of thousands of workers this year, the AI tool on your phone closed that gap and put a claim in front of a tribunal that would otherwise never have seen it. The same numbers say the tribunal was not built for the volume and some of the filings include facts the AI invented. Both are true. Action this week: If you are an Australian worker considering an unfair-dismissal claim, the Fair Work Commission's own guidance is at fwc.gov.au; the upcoming disclosure rule means you will need to state whether you used AI and whether you verified what it said, so check every precedent and date the AI gives you against the actual statute before you file. If you employ people anywhere, the surge in claims is going to land in your jurisdiction next; pull up your termination process and make sure your last-twelve-months dismissal paperwork would survive an AI-assisted claim that arrives faster and in higher volume than your usual cadence. If you read employment-law coverage, the SmartCompany write-up is the cleanest version of Justice Hatcher's testimony.

smartcompany.com.au: AI-generated unfair dismissal claims swamp Fair Work Commission

Australian Fair Work Commission caseload tops 55,000 as AI-generated unfair dismissal claims surge
smartcompany.com.au · May 29, 2026

» What to watch this week

Tomorrow's signal lands here.