Dissent Under the Microscope: How US Law Enforcement Is Surveilling Opposition to AI

The Philadelphia Bulletin: Where It Surfaced

  <p>In December 2025, the Delaware Valley Intelligence Center (DVIC) — a fusion center housed inside the Philadelphia Police Department — distributed a confidential alert warning of threats to AI data centers. The document, marked "for official use only," was circulated through the national fusion center network to state, local, and federal agencies.<sup><a href="#s1">[1]</a></sup></p>

  <p>The bulletin warned that "Domestic violent extremists (DVEs) are likely interested in targeting artificial intelligence (AI) data centers, posing a physical and cyber threat to infrastructure in the Philadelphia regional area." In the same document, the DVIC acknowledged "a lack of specific information on plans to target AI data centers in the Philadelphia area."<sup><a href="#s1">[1]</a></sup></p>

  <div>
    <span>Key Detail</span>
    The report listed "disruptive First Amendment activity" as an indicator of domestic violent extremism risk — placing constitutionally protected protest alongside actual threats in an intelligence assessment distributed nationwide.
  </div>

  <p>What the DVIC actually monitored: social media posts from AI opponents. Specific examples cited in the bulletin included an anonymous post on the Philly Anti-Capitalist blog titled "Butlerian Jihad Against AI" containing the language "ATTACK! ATTACK! ATTACK!" paired with a cartoon computer image, and a Facebook meme stating "I cannot escape the feeling that I am morally obligated to sabotage AI data center infrastructure."<sup><a href="#s1">[1]</a></sup></p>

  <p>Philadelphia civil rights attorney Paul Hetznecker called it a "dangerous attempt" to frame "legitimate, popular political concerns" as something more dangerous, warning it could chill legitimate community dialogue about data center impacts.<sup><a href="#s1">[1]</a></sup></p>

  <p>The Philadelphia Police Department's official response, delivered by Sergeant Eric Gripp, stated the fusion center "recognizes and respects the rights of individuals to lawfully express opinions" and that assessments are "designed to provide situational awareness, not to characterize lawful activity or constitutionally protected speech as criminal conduct."<sup><a href="#s1">[1]</a></sup> The bulletin itself contradicts this framing — it explicitly lists "First Amendment activity" as a threat indicator.</p>

  <h2>Not Just Philly: A National Infrastructure</h2>

  <p>The Philadelphia bulletin is not an isolated case. It is one node in a coordinated national surveillance effort. Over 1,000 pages of unpublished reports from the Department of Homeland Security, FBI, and regional intelligence fusion centers — obtained by WIRED — reveal a systematic shift to surveil what agencies internally call "anti-tech violent extremism."<sup><a href="#s2">[2]</a></sup><sup><a href="#s3">[3]</a></sup></p>

  <div>
    <span>Critical Finding</span>
    The term <strong>"anti-tech violent extremism"</strong> does not appear in any publicly available DHS or FBI domestic extremism report, guide, or handbook. It is an entirely new, unpublicized classification created within the intelligence community and circulated through internal channels only.<sup><a href="#s2">[2]</a></sup>
  </div>

  <p>The leaked documents show fusion centers across the country monitoring constitutionally protected events: in-person gatherings, town halls, public budget meetings, environmental protests where citizens raise concerns about noise pollution and utility costs, and even anti-Elon Musk "Tesla Takedown" rallies.<sup><a href="#s2">[2]</a></sup></p>

  <h2>Who's Watching</h2>

  <p>The surveillance apparatus spans federal, state, and local levels — connected through the national fusion center network.</p>

  <div>
    <div>
      <div>FBI / DHS</div>
      <p>Created the "anti-tech violent extremism" category. Produced internal threat assessments distributed to all 80+ fusion centers nationwide. Referenced in National Security Presidential Memorandum 7.<sup><a href="#s2">[2]</a></sup></p>
    </div>
    <div>
      <div>DVIC (Philadelphia)</div>
      <p>Distributed the December 2025 alert flagging "disruptive First Amendment activity" as a DVE indicator. Monitored anti-AI social media posts.<sup><a href="#s1">[1]</a></sup></p>
    </div>
    <div>
      <div>US Capitol Police ISB</div>
      <p>Intelligence Services Bureau issued an April 2026 intelligence note on data center opposition. Admitted it "is not investigating any data center-motivated threats to Members of Congress" — then warned opposition would increase.<sup><a href="#s4">[4]</a></sup></p>
    </div>
    <div>
      <div>NYC Intelligence Bureau</div>
      <p>New York City's Intelligence and Counterterrorism Bureau produced a report warning AI advances could spark "large-scale protests" and violent extremist activity.<sup><a href="#s3">[3]</a></sup></p>
    </div>
    <div>
      <div>Virginia Fusion Center</div>
      <p>A regional bulletin specifically warned about "anti-government extremists conducting surveillance around technology infrastructure."<sup><a href="#s3">[3]</a></sup></p>
    </div>
    <div>
      <div>80+ State Fusion Centers</div>
      <p>The national fusion center network — 80+ centers funded by DHS — serves as the distribution pipeline for these intelligence products.<sup><a href="#s5">[5]</a></sup></p>
    </div>
  </div>

  <p>This is not one rogue department. The Philadelphia bulletin was noteworthy for its explicit language about "First Amendment activity," but the surveillance infrastructure it feeds into is federal in scope and design.</p>

  <h2>What They're Monitoring</h2>

  <p>The leaked documents describe monitoring of activities that, in almost every case, are constitutionally protected:</p>

  <table>
    <thead>
      <tr>
        <th>Activity Monitored</th>
        <th>Constitutional Status</th>
        <th>Agency/Center</th>
      </tr>
    </thead>
    <tbody>
      <tr>
        <td>Social media posts critical of AI/data centers</td>
        <td>Protected speech (1st Amend.)</td>
        <td>DVIC, fusion centers</td>
      </tr>
      <tr>
        <td>Town hall attendance / public budget meetings</td>
        <td>Protected assembly (1st Amend.)</td>
        <td>Multiple fusion centers</td>
      </tr>
      <tr>
        <td>Environmental protests at data center sites</td>
        <td>Protected assembly (1st Amend.)</td>
        <td>Virginia, others</td>
      </tr>
      <tr>
        <td>"Extensive criticism of higher utility bills"</td>
        <td>Protected speech (1st Amend.)</td>
        <td>DVIC</td>
      </tr>
      <tr>
        <td>Online calls to boycott data centers</td>
        <td>Protected speech (1st Amend.)</td>
        <td>DVIC</td>
      </tr>
      <tr>
        <td>Photography of data center facilities</td>
        <td>Protected activity (general)</td>
        <td>Virginia fusion center</td>
      </tr>
      <tr>
        <td>"Tesla Takedown" anti-Musk rallies</td>
        <td>Protected assembly (1st Amend.)</td>
        <td>Multiple fusion centers</td>
      </tr>
    </tbody>
  </table>

  <p>Spencer Reynolds, senior counsel at the NAACP Legal Defense Fund, told WIRED that these intelligence reports are "part of a long tradition of agencies identifying protest or even simply having strong opinions as precursors to violence. Suspicious activity reports are incredibly unreliable, often about vague or innocent behavior, issued under permissive standards."<sup><a href="#s2">[2]</a></sup></p>

  <h3>The Capitol Police Contradiction</h3>

  <p>The US Capitol Police Intelligence Services Bureau's April 2026 intelligence note is particularly revealing. The document explicitly states: "The US Capitol Police is not investigating any data center-motivated threats to Members of Congress."<sup><a href="#s4">[4]</a></sup> Despite this admission, the report proceeds to warn that AI-related policies "are likely to continue drawing opposition, increasing potential concerns for public officials" and dedicates substantial space to threats against OpenAI CEO Sam Altman — who is not a government official.<sup><a href="#s4">[4]</a></sup></p>

  <p>The report identifies common public grievances — "possible government use of AI to spy on Americans," "environmental impacts," "rising energy costs," and "loss of jobs" — as reasons warranting intelligence monitoring, even when no actionable threat exists.<sup><a href="#s4">[4]</a></sup></p>

  <h2>The Playbook Has Been Used Before</h2>

  <p>Fusion center surveillance of constitutionally protected activity is not new. The current AI-focused monitoring fits a documented pattern stretching back nearly two decades.</p>

  <div>
    <div>
      <span>2007–2009</span>
      <p>Fusion centers targeted Muslim civil rights organizations, anti-war protest groups, and lobbying organizations. A leaked 2009 document showed it was "imperative for law enforcement officers to report" the activities of these groups.<sup><a href="#s6">[6]</a></sup></p>
    </div>
    <div>
      <span>2012</span>
      <p>The Senate Homeland Security Committee's permanent subcommittee on investigations found fusion centers produced "little information of value" despite lavish DHS funding, documented "waste, mismanagement, and incompetence," and uncovered serial civil liberties violations and "excessive secrecy."<sup><a href="#s6">[6]</a></sup></p>
    </div>
    <div>
      <span>2020</span>
      <p>A fusion center contractor breach exposed "hundreds of thousands of sensitive records from the FBI, DHS, and other law enforcement agencies."<sup><a href="#s5">[5]</a></sup></p>
    </div>
    <div>
      <span>2021–2023</span>
      <p>Atlanta deployed 124 surveillance cameras per 1,000 residents — exceeding most cities outside China — and used networked AI surveillance against Forest Defender activists opposing Cop City. One activist, Manuel "Tortuguita" Paez Terán, was shot and killed by a Georgia State Patrol officer in January 2023.<sup><a href="#s7">[7]</a></sup></p>
    </div>
    <div>
      <span>2024</span>
      <p>Fusion centers in Ohio, California, and South Carolina surveilled pro-Palestine student protesters. Ohio State Police arrested 36 students within two hours of a fusion center alert. The FBI provided electronic device warrants to campus police at Cal Poly Humboldt. Universities partnered with fusion centers to track student IP addresses and WiFi connections.<sup><a href="#s8">[8]</a></sup></p>
    </div>
    <div>
      <span>Dec 2025 – June 2026</span>
      <p>The "anti-tech violent extremism" framework emerges. Philadelphia DVIC bulletin, 1,000+ pages of leaked federal documents, Capitol Police intelligence note, NYC and Virginia fusion center reports — all targeting AI data center opposition.<sup><a href="#s1">[1]</a></sup><sup><a href="#s2">[2]</a></sup></p>
    </div>
  </div>

  <p>The Brennan Center for Justice documented in 2022 that fusion centers have "repeatedly targeted minority communities and protest movements under the guise of counterterrorism or public safety," and that DHS has "failed to ensure that they have used these resources appropriately." The centers continue to operate with "little oversight or public accountability."<sup><a href="#s5">[5]</a></sup></p>

  <h2>The Scale of Opposition They're Watching</h2>

  <p>The intelligence agencies are monitoring opposition that represents mainstream American opinion — not a fringe.</p>

  <div>
    <div>
      <span>70%</span>
      <span>Americans opposing local data centers (Gallup, March 2026)</span>
    </div>
    <div>
      <span>$98B</span>
      <span>In data center projects stalled by activists (Q2 2025)</span>
    </div>
    <div>
      <span>5×</span>
      <span>More Americans concerned than excited about AI (Pew, 2025)</span>
    </div>
  </div>

  <p>A Gallup poll conducted March 2–18, 2026, found seven in ten Americans oppose AI data center construction in their local area, with 48% <em>strongly</em> opposed. This represents a massive jump from the 47% who objected in late 2025. Opposition is now so intense that respondents would rather live near a nuclear power plant than a data center.<sup><a href="#s9">[9]</a></sup></p>

  <p>The grassroots movement spans ideological lines. In February 2026, nearly 200 activists gathered at a church in Richmond, Virginia to oppose data centers, then marched to the state capitol. State delegate John McAuliff told them: "You're getting a sh-t deal." Activists have successfully defeated data center projects in Wisconsin, Oklahoma (where the Muscogee Nation rejected a hyperscale proposal), and other states.<sup><a href="#s10">[10]</a></sup></p>

  <blockquote>
    Politicians who choose to do the bidding of Big Tech at the expense of hardworking Americans will pay a huge political price.
    <cite>— Brendan Steinhauser, conservative strategist<sup><a href="#s10">[10]</a></sup></cite>
  </blockquote>

  <p>Georgia residents in counties hosting data centers experienced six rate hikes between 2023 and 2025.<sup><a href="#s10">[10]</a></sup> Half of opponents cite excessive resource consumption — 18% specifically mentioning water use and 18% energy use — while 16% cite pollution concerns including noise, air, and water contamination.<sup><a href="#s9">[9]</a></sup></p>

  <p>The surveillance apparatus is, in effect, monitoring mainstream political opposition held by a supermajority of Americans.</p>

  <h2>The Chilling Effect on AI Dissent</h2>

  <p>The legal concept of a "chilling effect" occurs when individuals seeking to engage in First Amendment-protected activity are deterred from doing so by government action. Courts have recognized that the broader or vaguer surveillance authority becomes, the greater its deterrent effect on free speech — and that the mere existence of such a program can chill expression, regardless of how it is applied in practice.<sup><a href="#s11">[11]</a></sup></p>

  <h3>How It Works in Practice</h3>

  <p>The surveillance infrastructure identified in the leaked documents creates chilling effects through several mechanisms:</p>

  <p><strong>Social media monitoring deters online speech.</strong> When citizens know that agencies are scanning social media for "anti-tech" sentiment and compiling intelligence bulletins from public posts, rational self-censorship follows. The DVIC bulletin demonstrated this directly — anonymous blog posts and Facebook memes were compiled into a counterterrorism-adjacent intelligence product.</p>

  <p><strong>Fusion center alerts deter attendance at public forums.</strong> The leaked documents show monitoring of town halls, public budget meetings, and county hearings. If attending a public meeting about your utility bill puts you in an intelligence database, some people will stay home.</p>

  <p><strong>The "domestic violent extremist" label itself deters participation.</strong> By categorizing AI opposition alongside actual violent extremism, the intelligence products stigmatize legitimate civic engagement. A community member worried about water usage near a proposed data center now shares a threat category with white supremacists and anarchists.</p>

  <p>The ACLU has documented that fusion center operations without guidelines and standards constitute "an assault on citizens' rights to privacy, free speech, and freedom of assembly," with indiscriminate investigations producing "a chilling effect on free speech and association."<sup><a href="#s6">[6]</a></sup></p>

  <h3>The Broader AI Surveillance Feedback Loop</h3>

  <p>There is a grim irony in using AI-powered surveillance tools to monitor people who oppose AI. The ACLU of Massachusetts documented that over 80 police departments in that state alone have adopted AI-powered license plate readers with no state regulation, that companies like BriefCam provide machine learning to "automatically identify and follow individuals across multiple camera streams," and that vendors including Axon, Motorola, and Genetec offer "AI-powered predictive policing" tools that consolidate surveillance data with facial recognition.<sup><a href="#s12">[12]</a></sup></p>

  <p>As Jordan Harmon, a Muscogee Nation policy specialist, noted: data center infrastructure supports "AI facial recognition to surveil and police people, including Indigenous people, who are also being detained by ICE."<sup><a href="#s10">[10]</a></sup> The infrastructure being protested enables the surveillance of the protesters.</p>

  <h2>Analysis: What This Means</h2>

  <h3>Is this widespread or confined to Philadelphia?</h3>

  <div>
    <span>Verdict</span>
    <strong>This is a coordinated national effort, not a local anomaly.</strong> The Philadelphia DVIC bulletin was one product among 1,000+ pages of federal intelligence documents. The FBI, DHS, Capitol Police ISB, NYC Intelligence Bureau, Virginia fusion centers, and the entire national network of 80+ fusion centers are involved. The Philadelphia case is notable only because its language about "First Amendment activity" was unusually explicit — not because it is unique in scope.
  </div>

  <h3>What does this mean for dissenting opinions against AI?</h3>

  <p>Three concrete consequences:</p>

  <p><strong>1. Dissent is being pre-criminalized.</strong> The intelligence framework treats opposition to AI infrastructure as a <em>precursor</em> to violent extremism rather than constitutionally protected civic participation. This is not post-hoc investigation of actual crimes — it is proactive surveillance based on viewpoint. When "extensive criticism of higher utility bills" appears in a counterterrorism document, the line between political dissent and security threat has been erased.</p>

  <p><strong>2. A supermajority opinion is being treated as a fringe threat.</strong> 70% of Americans oppose local data centers. Five times as many Americans are concerned about AI as are excited by it. The surveillance apparatus is monitoring views held by the mainstream, using a framework designed for detecting genuinely rare violent extremism. This is a category error with constitutional implications.</p>

  <p><strong>3. The surveillance infrastructure and the thing being protested are the same system.</strong> AI powers the surveillance tools (facial recognition, predictive policing, social media monitoring) used to watch AI critics. Data centers host the computing power for these tools. Opposing the infrastructure that enables your own surveillance puts you on a watchlist maintained by that same infrastructure. This feedback loop has no precedent in previous protest movements.</p>

  <div>
    <span>The Strongest Counter-Argument</span>
    Law enforcement has a legitimate interest in preventing actual attacks on critical infrastructure. Genuine threats exist: the April 2026 alleged Molotov cocktail attack on Sam Altman's home,<sup><a href="#s3">[3]</a></sup> and some online posts monitored by the DVIC did reference explosives and electromagnetic pulse weapons.<sup><a href="#s1">[1]</a></sup> The question is not whether threats should be investigated — they should — but whether the intelligence framework treats the <em>opinion itself</em> as the threat indicator, which these documents plainly do. Surveilling specific credible threats is policing. Categorizing an entire viewpoint as pre-criminal is something else.
  </div>

  <h3>Historical Pattern, New Target</h3>

  <p>This follows a documented cycle. Fusion centers targeted Muslim communities after 9/11, anti-war protesters during the Iraq War, racial justice activists during the BLM movement, environmental activists during pipeline protests, and pro-Palestine students in 2024. In each case, constitutionally protected activity was reframed as a security threat, intelligence products were produced with minimal oversight, and the surveillance expanded until exposure forced reforms — which were then eroded before the next cycle began.</p>

  <p>The AI opposition surveillance is the latest iteration of this pattern, with one key difference: the technology being protested is also the technology powering the surveillance. That structural entanglement — where the object of dissent is also the instrument of monitoring — creates a self-reinforcing system that previous cycles did not face.</p>

  <p>The leaked documents make clear that US law enforcement has decided, at the federal level, that opposition to AI is a domestic security concern. Whether you believe that framing is justified or terrifying depends on how much you trust the institutions making that determination — institutions that, as the Senate found in 2012, have produced "little information of value" while committing "serial violations of civil liberties protections."<sup><a href="#s6">[6]</a></sup></p>