Censorship-industrial complex gunning for AI, spending-bill goodies, House Republicans warn

Congress must block "needless regulation" of artificial intelligence, a "new frontier" for the so-called censorship-industrial complex after it badgered and colluded with Big Tech to suppress disfavored narratives on elections, COVID-19, climate change and pardoned convict Hunter Biden's abandoned laptop, a House Judiciary Committee panel says.

The Weaponization of the Federal Government Subcommittee's interim staff report calls to end funding of "censorship-related research" and foreign collaboration on AI censorship, take federal hands off "private algorithm or dataset decisions" related to purported misinformation or bias, and ban any regulation "that gives the government coercive leverage."

The panel documented the National Science Foundation's funding of "AI-powered speech-monitoring tools that could enable the mass censorship of American speech" in February, prompting Democrats to retort that then-candidate Donald Trump was a bigger threat.

"AI companies, aware of the power that federal regulators have over their future," even let the Biden administration see their new models in development, the report says, citing a National Institute of Standards and Technology announcement this summer. 

The report quotes the Biden White House touting its pressure to "advance equity," stop "algorithmic discrimination" and "mitigate the production of harmful and biased outputs" through AI, which the panel said gives the feds the means to "monitor, suppress, and ultimately censor."

Anti-censorship lawmakers are keeping their eyes peeled for remnants of the outgoing regime as they negotiate a spending bill to avert a partial government shutdown, with a revised version failing Thursday night. 

The Global Engagement Center, a State Department-hosted "interagency partnership" that funds advertising boycotts against conservative publishers and promotes anti-populist internet games, unexpectedly showed up in the 1,500-page continuing resolution that went down in flames after opposition from President-elect Trump and his Terminator-like ally Elon Musk.

The yearlong extension of GEC, which was set to formally shutter next week due to expiring authorization, outraged Republicans including Georgia's Andrew Clyde, who said "Christmas Cramnibus is a vote for taxpayer-funded censorship," and Indiana's Jim Banks, who demanded answers from State after GEC tried to tie him to "Russian influence operations."

House Speaker Mike Johnson's office defended the extension to Washington Examiner reporter Gabe Kaminsky, who first spotted it, saying the Louisiana Republican "has killed multiple efforts to pass a 5-year reauthorization" and the provision gives Trump "maximum ability and authority to determine how to handle the office, its authorities, and funding."

Other routine targets of disinformation police and their federal funders cried foul at CR provisions including a two-year extension of the Pandemic and All-Hazards Preparedness Act, which grants sweeping emergency powers and immunizes "countermeasures" providers.

Vaccine-injury advocacy group React19 alerted Just the News to a provision that would limit the powers of presumptive Secretary of Health and Human Services Robert F. Kennedy Jr. over the "Vaccine Injury Table" for any vaccine recommended for children and pregnant  women by the Centers for Disease Control and Prevention.

The provision would "also destroy" the Vaccine Injury Modernization Act (HR 5142) "by explicitly precluding the transfer" of React19's cases under the Countermeasures Injury Compensation Program — which makes compensation nearly impossible to recover — to the friendlier Vaccine Injury Compensation Program, legal affairs director Chris Dreisbach wrote.

Government misuse of AI, or just sloppiness, doesn't just worry Hill Republicans. New York City's Education Department pulled a $1.9 million proposal for an AI-powered "reading tutor" hours before a Wednesday vote, after City Comptroller Brad Lander warned it "needed clearer policies and student privacy safeguards," Chalkbeat reported.

From NSF and GEC to the FBI, Cybersecurity and Infrastructure Security Agency and White House's profane harangues against insufficiently compliant tech executives, "individuals at every level … have used veiled threats of retaliation to coerce social media companies to silence the voices of American citizens," the GOP-led Weaponization report says.

It often refers back to the February hearing on AI censorship, where witnesses such as Foundation for Individual Rights and Expression CEO Greg Lukianoff and former Intercept reporter Lee Fang warned that "regulatory panic" could give a minority power over "even questions" and the U.K. already used AI narrative surveillance during COVID and elections.

"Armed with AI-powered content moderation tools," some of them publicly funded, "Big Tech can more fully and quickly comply with the government’s censorship demands," the report says. 

It warned that a "successor" to the Department of Homeland Security-conceived Election Integrity Partnership, a private consortium of more than 100 staff to mass-report election narratives for removal, throttling and "soft-blocking" in 2020 and 2022, "could monitor and flag tens of millions of Americans’ election-related posts for censorship."

Government involvement in an AI model's "combination of training data and machine learning algorithms" could teach it to ignore certain information so that users requesting it draw a blank, or "censor certain outputs regardless of the training data by manipulating the model’s machine learning algorithm," the report says.

Switching to the economy and national security, the GOP-led panel argues for a light regulatory touch to ensure programmers stay stateside and away from foreign adversaries. The report hints at First Amendment litigation from AI creators if Congress or agencies restrict "expressive uses of AI," impeding their editorial decisions on training data and user prompts.

What the outgoing administration calls "frameworks" and "blueprints" are actually veiled threats to "censor your AI models, or else," the report says. 

Starting in 2021, when the nominally neutral NIST developed an "AI Risk Management Framework" with companies to "incorporate trustworthiness considerations" into all stages of AI, the White House and agencies repeatedly leaned on and secured private concessions, the panel said. A national security memorandum this fall assigned NIST's AI Safety Institute with creating guidance for developers on "malicious use of AI to generate misleading videos or images [] of political or public figures," plausibly including satirical memes.

President Biden's own executive orders ditched the pretense of voluntary cooperation, ordering developers to share "dual-use foundation models" with the feds and tasking the Justice Department with policing "algorithmic discrimination," the report says.

Google parent Alphabet's attempt to comply with White House marching orders, including consulting activists to mitigate the supposed harms of its Gemini AI, may have played a role in Gemini's "historically inaccurate outputs" such as refusing to show white people, it says.

"The federal government intends to become the AI gatekeeper, ensuring that only models complying with its censorship demands are released," according to the panel.