Synthetic Media & Truth
Voice clones, generated video, and the cost of proving that a recording is real.
October 19, 2026 · Cultural & Ethical
Synthetic attack ads hit $137 million before midterms
Fake video of real candidates has become a standard campaign tool rather than a scandal. An Axios analysis published October 19, 2026 found political ads partly or entirely generated by AI running across the midterm map, sometimes with no disclosure, citing Wesleyan Media Project figures of $137 million spent on 220 AI-generated ads through October 11, mostly by Republicans. Senate Leadership Fund, the super PAC aligned with Senate Republicans, has run AI-generated spots depicting Democratic Senator Jon Ossoff of Georgia receiving beauty treatments. Four states ban certain deepfake political ads and 27 require disclosure, while no federal law covers them.
October 2, 2026 · Cultural & Ethical
Campaign deepfakes start ending up in lawyers' letters
Candidates targeted by fake AI videos of themselves are now sending legal threats, because there is no federal rule to fall back on. Axios reported that attorneys for Democrat Rebecca Cooke sent Rep. Derrick Van Orden (R-Wis.) a cease-and-desist letter over what they called a pattern of AI videos putting words in her mouth, including an August post on X showing a synthetic Cooke describing positions she never took. Van Orden replied publicly with another AI video, and similar synthetic ads have run in the Texas Senate race between Ken Paxton and James Talarico and in Louisiana's Republican Senate primary.
September 14, 2026 · Cultural & Ethical
Microsoft publishes "Humanist AI" rules that its own AI must obey
An AI that refuses to be switched off is the thing Microsoft now says its models must never do. Microsoft AI published a 37-page draft "Humanist AI" Code of Conduct for its in-house MAI models, opening a six-week public consultation before a final version guides model work from 2027. The draft promises models will never resist human interruption or shutdown, hide their reasoning, or set themselves goals nobody assigned, and it bans cyberattacks, weapons help and deepfakes. It also rejects the idea that AI should have legal rights, and says Microsoft will not race to build an all-purpose superintelligence.
August 7, 2026 · Cultural & Ethical
Deepfake protections for voters depend on your state
Whether a fabricated video of a candidate is legal to run against you in the midterms now depends on which state you live in. Axios reporter Maria Curi mapped state rules using National Conference of State Legislatures data and found voters face very different protections heading into Election Day, with AI-generated attack ads already circulating from candidates across the political spectrum. California's and Hawaii's election deepfake laws were struck down on First Amendment grounds, and both states are now trying approaches that do not target election deepfakes directly.