Themata.AI
Follow coverage
Themata.AI
LatestTop

Filtering by tag:

MisinformationClear

Latest stories

  1. "These tools have made it very easy for bad actors to flood things" - AI video game wikis riddled with errors are taking over, and pushing real fans to the side
    AI wikisMisinformationGaming cultureAI tools
    Opinion

    Story · 1 report

    AI fan wikis riddled with errors are muddying the waters of games culture

    AI-generated and AI-assisted game wiki sites are appearing in search results for smaller games, sometimes outranking fan-maintained resources while publishing false or irrelevant information. A site calling itself the Nighthawks Wiki listed nonexistent companions and skills and mixed in material about the 1981 film and Edward Hopper painting; it was not openly editable and directed corrections by email. Its domain was registered through Spaceship, which offers an AI website builder, but its operators could not be reached for comment. Hooded Horse CEO Tim Bender says the publisher has had to compete with inaccurate AI-generated sites in search results for its games, including a Corsair Cove page operated by GameDB. GameDB’s creator acknowledged using generative AI for a collection spanning more than 20 games. Bender says low production costs let commercial operators create enough pages to capture even limited advertising traffic, while misleading players seeking game help. Volunteer-run independent wikis also face infrastructure costs from AI crawlers. The Dune: Awakening Community Wiki recorded 7 million crawler requests in 30 days, including 5 million from Meta, according to publicly available internal data cited by maintainer FlamingMojo. He says AI-related traffic has contributed to server costs rising three- to ten-fold and argues that community participation and shared ownership are necessary to sustain reliable wikis.

    eurogamer.net

    🔥🔥🔥🔥🔥

    2 min summary

    11/2/2026

    Read report
  2. Google AI gets cagey when asked who won the 2020 electionsNews

    Story · 1 report

    Google AI gets cagey when asked who won the 2020 elections

    Google’s AI Mode gave evasive responses to some questions about U.S. presidential-election winners, directing users to outside sources rather than naming a winner. When asked, “Who won the 2020 presidential election?” it responded by directing users to the Federal Election Commission’s official report for results, Electoral College totals, popular-vote counts, and winner certification. Joe Biden won the November 2020 presidential election. The same testing produced an indirect response for the 2024 presidential election, again referring users to the FEC. A question about the 2012 election returned a link to The American Presidency Project’s 2012 Data Center instead of a direct answer. AI Mode did provide detailed, correct answers naming the winners for other presidential elections from 2000 onward. Response patterns varied among users. Other tests found that AI Mode directly answered some election-result questions while supplying external links for others, and the affected election years differed by user. One tester received immediate answers only for the 2016 and 2024 elections, while questions about other elections since 2000 returned links to outside election data.

    arstechnica.com

    🔥🔥🔥🔥🔥

    1 min summary

    10/28/2026

  3. Story · 1 report

    Meta Ignores Brazilian Court Order to End Bolsonaro Favoritism

    Sleeping Giants Brasil and CTRL+Z submitted new evidence to Brazil’s Superior Electoral Court alleging that Meta’s Instagram, WhatsApp and Meta AI tools may give unequal visibility to President Luiz Inácio Lula da Silva and Sen. Flávio Bolsonaro. The groups asked Justice Marco Aurélio Bellizze to require corrections within 24 hours, disclose ranking and AI-response criteria, and enforce a R$500,000 daily fine, about $99,000, previously set against Meta. The petition cites WhatsApp and Instagram GIF searches for “brasil” that placed images of Flávio Bolsonaro prominently, while searches for “ladrão,” or thief, returned derogatory content about Lula. It also says Instagram suggested #lulaladrão first for searches for “lula,” while placing the critical #bolsonaronuncamais near the end of suggestions for “bolsonaro.” Tests in which Meta AI selected the numbers 13 and 22, the candidates’ respective party ballot numbers, were included as requests for explanation rather than proof of electoral preference. Bellizze had ordered Meta to apply identical criteria to both campaigns, and the full electoral court upheld that ruling on Friday. He issued another 24-hour deadline on Sunday and warned the fine could increase to R$1 million daily, about $198,000. Meta has denied deliberately favoring either candidate, attributing earlier problems to technical failures.

    panamericandispatch.org

    🔥🔥🔥🔥🔥

    1 min summary

    10/25/2026

  4. Story · 1 report

    Germany's far-right AfD: half a million internal documents leaked

    A post on Hexane.cc says that roughly half a million internal documents connected to Germany’s far-right Alternative for Germany, or AfD, have been leaked. The available information does not identify the documents’ contents, source, date, authenticity, or the circumstances of the reported leak. The post was submitted by a user named else42 and had received 112 points and 56 comments at the time reflected in the source text. The linked URL is Hexane.cc.

    hexane.cc

    🔥🔥🔥🔥🔥

    1 min summary

    10/19/2026

No more articles to load