When AI Decides You're a Threat — Brad Carson

What It Covers

Brad Carson, former U.S. Congressman and Department of Defense official, examines how AI reshapes military targeting, autonomous weapons accountability, deepfake liability, and AI regulation. He argues against technological fatalism, advocates for mandatory frontier model testing, US-China diplomatic engagement on AI, and congressional oversight to prevent informal Silicon Valley capture of AI policy.

Key Questions Answered

• •AI Targeting & False Positives: Military AI systems now assign probabilistic threat scores — a 0.73 probability someone is a Hamas combatant — replacing categorical combatant/civilian distinctions. Commanders accept pre-acknowledged false positive rates without understanding how scores are derived. Carson argues this opacity makes accountability impossible: you cannot court-martial an algorithm, fundamentally breaking centuries of war crimes accountability frameworks built around identifiable human decision-makers.
• •Regulatory Capture vs. Regulatory Vacuum: Carson counters the "regulatory capture" argument by pointing out the current alternative is worse — informal networks of wealthy Silicon Valley figures shaping AI policy through political contributions and influence. He uses the SEC analogy: imperfect regulated oversight still outperforms unaccountable private control. The falsifiability problem with capture arguments is that they justify doing nothing while concentrated power grows unchecked.
• •Frontier Model Liability Framework: AI developers should bear primary but not total liability for harmful outputs, following existing product liability law. Companies that knowingly include child sexual abuse material in training data, fail to engineer out suicide-encouraging behavior, or ignore foreseeable misuse patterns should face legal consequences. The tort system's "entity most capable of avoiding risk" principle directly applies to the five frontier labs spending hundreds of billions.
• •AI Anthropomorphization as Legal Threat: Tech companies are strategically arguing AI outputs constitute First Amendment-protected speech to block all regulation. Carson identifies this as a deliberate legal strategy after losing legislative battles — not a genuine philosophical position. Treating AI as a product under existing product liability law provides a coherent alternative framework, allowing regulation of harmful outputs without the constitutional barriers that human speech protections would create.
• •Chip Control as AI Arms Race Exit: The US and allied nations control the entire semiconductor supply chain for frontier AI — NVIDIA, ASML, Japanese photoresist manufacturers. This concentration means the US could halt other nations' frontier AI development without firing a shot. Carson argues this leverage makes AI fatalism factually wrong: unlike nuclear weapons, the physical infrastructure required for superintelligent AI remains geographically and commercially controllable by Western governments.
• •Congress Needs Restored Tech Advisory Infrastructure: Newt Gingrich eliminated the congressional Office of Technology Assessment in 1994, leaving legislators with only 17 minutes daily for issue education and no independent technical brain trust. Carson advocates restoring a congressionally chartered, government-funded technical advisory body insulated from both industry lobbyists and philanthropist-influenced nonprofits. Without it, AI policy defaults to whichever well-funded informal network has the most political access.