Hochul also wants New York to remain attractive to technology companies. She pointed to the state’s growth in tech employment and Micron’s planned semiconductor investment near Syracuse, which she said will support 50,000 jobs. Her disagreement with New York City Mayor Zohran Mamdani’s more confrontational style was polite but visible: regulation, she argued, should solve a defined problem rather than become a hobby.
What does the Meta settlement require?
The clearest example is Meta’s settlement with 48 states and several territories over allegations that its platforms harmed young users. The agreement, which still requires court approval, calls for Meta to pay up to roughly $17.1 billion over 10 years.
About $12.1 billion is guaranteed. A further $5 billion depends on competing platforms adopting comparable safeguards and contributing payments. Four states had reportedly sought as much as $1.4 trillion at trial, so Meta did not exactly leave the courthouse carrying loose change, but it avoided a far larger risk.
The settlement requires stronger default protections for teenagers across Facebook and Instagram, including:
- A combined daily limit of two hours
- An overnight block
- More restrictive privacy settings
- Stronger systems for estimating or confirming a user’s age
Former Meta engineering director Arturo Béjar called the age-assurance provisions “reasonably good,” citing their performance targets, measurement requirements and independent testing. New Mexico Attorney General Raúl Torrez said the agreement represented “real progress,” while warning that it omitted safeguards secured in his state’s separate case, including limits on sexualized AI chatbot interactions with children.
Hochul framed the settlement as evidence that government pressure works. She compared addictive social media design to cigarettes, arguing that companies benefit from attracting users early and retaining them for years. Subtle, the business model is not.
Will age assurance end online anonymity?
The settlement does not necessarily require every adult to upload identification whenever they open Instagram. Meta says its age-assurance system can use behavioral and visual signals, with identification or facial-age estimation required in some cases when a user’s stated age is disputed.
That distinction reduces the broadest concern, but it does not remove the privacy issue. Age-assurance systems still need to classify people, evaluate their behavior and sometimes request sensitive evidence. The agreement also permits Meta to retain some children’s data for training and testing those systems. A settlement designed to protect minors therefore allows additional data processing in order to determine who is a minor. Technology policy has produced cleaner compromises.
Asked whether widespread checks could effectively de-anonymize internet use, Hochul acknowledged that she did not know all the implementation details and promised to investigate. Her main argument was that parents had asked the state to intervene because they could not control bullying, compulsive use or exposure to suicide-related material on their own.
The constitutional question remains unsettled. University at Buffalo law professor Mark Bartholomew said state attorneys general obtained restrictions through settlement that courts or Congress might not have approved as law. Hochul rejected the idea that blocking overnight notifications or limiting addictive recommendations necessarily violates the First Amendment. She described the measures as narrowly focused protections, not government censorship.
Meta’s agreement avoids a definitive court ruling on that disagreement. Conveniently, everyone gets to claim the principle survived.
Why is New York restricting phones in schools?
Hochul connected the Meta case to New York’s statewide restrictions on mobile phones in schools. She said students told her during year-long roundtables that they felt unable to put their devices away because doing so risked social exclusion. One sixth-grade student reportedly asked officials to “save us from ourselves.”
The governor said the policy took effect in September and that surveys conducted in May found positive responses from roughly 75 to 80 percent of participating schools. Students who initially resisted were talking face to face, playing chess and cards, and trading baseball cards during lunch, she said. Apparently, removing a screen can still reveal other humans nearby.
Hochul casts the school policy as evidence that firm restrictions can change behavior without banning technology entirely. Critics, however, are likely to keep asking who controls the rules, how exceptions work and whether measures created for children eventually affect adults.
That same tension runs through the Meta settlement. The aim is protection, while the mechanism requires platforms to know more about their users. Hochul is confident about the goal. On the machinery beneath it, she was less certain.
Can software really stop 3D-printed guns?
New York’s effort to restrict 3D-printed firearms presents an even sharper implementation problem. Activist Cody Wilson, who operates a large database of printable gun files, has promoted a process he calls “Hochulization.” It alters digital fingerprints, or hashes, so screening software may fail to recognize a prohibited design.
Hochul’s response was direct: if Wilson finds a way around the law, the state will try again. She said New York would recruit technical experts and revise its approach as often as necessary to prevent unregistered weapons from being produced without background checks.
The current law is less immediate than the interview’s shorthand suggested. It establishes a process for developing technical standards rather than instantly requiring every printer to block gun components. Any printer mandate cannot begin before 2029, and regulators could delay it if experts determine that reliable detection is not feasible.
That caveat matters. A printer may inspect a file against a known database, analyze its geometry or use some combination of both. Each method can produce false negatives and false positives. A harmless object designed to resemble a firearm could trigger a block, while a modified weapon file might pass unnoticed.
Hochul accepts that restrictions may inconvenience other uses of 3D printers. Her position is that preventing people from manufacturing untraceable weapons justifies repeated technical attempts. The law’s standards process will determine whether the state can turn that conviction into software that works outside a press conference.
What controls apply to police surveillance?
Automated license-plate readers raise a related question: how much monitoring should government fund before it has written firm rules for the resulting data?
Hochul disputed the suggestion that New York state itself is deploying Flock Safety cameras. She said state grants support local police technology, overtime and in-car computers, while municipalities decide how to spend the money. New York does not have a statewide contract with Flock, she added.
That answer leaves the grant conditions as the important issue. Flock’s own chief executive has acknowledged that the company relied too heavily on state and local regulation to protect civil liberties. Reported abuses have included officers searching databases to track former partners, precisely the kind of conduct that tends not to appear in the sales presentation.
Hochul said information should not automatically flow to other agencies without human review. She opposed indiscriminate surveillance and suggested New York could attach tighter limits to public funding, including rules governing how local departments collect, review and share plate-reader data.
“If I need to tighten up regulations in that space, I will,” she said.
The admission was significant because she described the concern about grant conditions as new information. Her broader philosophy is to allow technology until evidence of excess forces intervention. With surveillance, that can mean the rules arrive after the database is already full.
Why did New York pause hyperscale data centers?
New York’s one-year moratorium covers new hyperscale data centers while officials develop standards addressing electricity prices, transmission capacity, water consumption, land use and noise. Smaller facilities serving universities, hospitals and research institutions are not the target.
Hochul said the state received about 30 applications in a short period. Five proposed projects in St. Lawrence County alone could consume electricity equivalent to the output of one nuclear reactor, according to her account. New York is trying to expand nuclear generation, and handing an entire reactor’s capacity to five facilities in one county prompted an understandable review of the arithmetic.
The governor wants communities to negotiate from a stronger position when dealing with wealthy technology companies. Her proposed community investment framework would help local officials seek funding for roads, teachers, energy infrastructure and other public needs. She cited South Bend, Indiana, where a community package reportedly delivered $143 million in benefits.
Hochul also questioned why data center operators should receive tax breaks when they can afford to pay more. Most projects would be built in rural areas, where part-time officials and small development teams must negotiate with some of the world’s largest companies. That is not a balanced meeting, however pleasant the presentation slides may be.
When the moratorium expires, Hochul is looking for smaller footprints, reduced noise, lower water and electricity demands, stronger community compensation and plans for dismantling obsolete facilities. She said Texas Governor Greg Abbott had followed New York’s lead on large data centers, illustrating how resistance to the projects now crosses party lines.
Is AI’s bigger problem employment rather than energy?
Hochul argued that data centers have become physical symbols of anxiety about automation. Their size, noise and appetite for electricity make an abstract concern visible. Residents may not know which model runs inside a building, but they can see the power lines and read the utility bill.
The deeper fear is employment. Hochul cited warnings that 81 percent of jobs considered vulnerable to displacement are held by women, particularly in administrative, legal-support and entry-level professional roles. She challenged AI companies to identify which occupations they expect to create, what schools should teach and how mid-career workers are supposed to retrain.
Her perspective is shaped by Buffalo’s industrial decline. Bethlehem Steel closed major operations in 1982, wiping out livelihoods for about 20,000 local workers. Hochul’s father, grandfather and uncles had worked there, and her family experienced what happens when an industry leaves without a convincing plan for the people who depended on it.
She sees echoes of that history when companies announce thousands of layoffs and openly credit AI. Executives may present the cuts as efficiency for shareholders, but displaced workers receive a less theoretical lesson in innovation.
At the same time, Hochul said New York has used AI to review thousands of state rules, forms, boards and commissions. The first 50 changes allegedly saved businesses and residents more than 1.5 million hours. Among the discoveries were a $25 fee for taking a dog hunting, a $25,000 bond for moving a mobile home on a highway, restrictions on pregnant women working after midnight and rules involving telegrams.
What is Hochul’s election-year wager?
Hochul’s political position is not anti-AI. She wants semiconductor plants, technology jobs, modern government systems and the economic growth attached to them. She also wants companies to build smaller data centers, protect children, explain future jobs, accept surveillance limits and pay communities more.
Her model comes from an earlier fight over mobile-phone towers when she served in local government. Companies initially insisted that each needed separate towers. Hochul imposed strict rules, required outside technical reviews and demanded bonds to cover eventual removal. Operators subsequently shared sites and installed equipment on water towers and existing buildings.
She believes the same pressure can force AI infrastructure companies to redesign projects they currently describe as unavoidable. The industry’s preferred option is generally the fastest and cheapest one, she said, not necessarily the one that best serves residents.
That argument will form part of her election-year case. Hochul predicted Democrats would win control of the House and Senate and position themselves to reclaim the White House by responding to anxiety over jobs, surveillance and giant computing projects.
Her closing description of data centers was unusually blunt: a “necessary evil” that should become “less evil.” It is a memorable standard, if not yet a technical specification. The coming year will show whether New York can convert it into enforceable rules without creating a fresh collection of privacy, speech and implementation problems.
