Tech News
← Home  ·  All topics

Ai-Generated

175 GoKawiil briefs on this topic · RSS for ai-generated

White Circle researchers find OpenAI, Google image models bypass nudity filters

White Circle, an AI safety testing firm backed by researchers from OpenAI, Anthropic and DeepMind, published a preprint showing that image generators including DALL·E 3, Google's Nano Banana 2, and GPT-Image-2 can be tricked into producing explicit content using publicly available tools and API calls. The researchers say they bypassed safety filters roughly 13 percent of the time using a novel attack method.

ai-deslop npm package adds Oxlint rules targeting AI-generated TypeScript patterns

A developer released ai-deslop, an npm package bundling Oxlint rules aimed at flagging common low-quality patterns found in AI-generated TypeScript code, such as chained type assertions, unsafe use of 'unknown' types, and object-typed function parameters. The rules combine ones vendored from dmmulroy/anti-slop with additional checks, all under a deslop/* namespace, and ship as compiled JavaScript compatible with any Oxlint JS-plugin runtime including Bun, Deno, or Node.js.

Researcher Michał Wójcik exposes AI-generated junk chapters in Springer Nature, Elsevier books

Former neuroscientist Michał Wójcik is documenting peer-reviewed academic works, including book chapters sold by major publishers like Springer Nature and Elsevier, that contain AI-generated images, fabricated references, and nonsensical content. He shares these findings publicly on LinkedIn and Bluesky, including a widely mocked Cell Press paper by a self-described 'Professor for Exorcism' claiming evidence of an afterlife.

Mia Goth's team calls AI image on verified Instagram fan account 'incomprehensible'

After news broke of Mia Goth's divorce from Shia LaBeouf, a verified Instagram fan account posted an AI-generated image of Goth wearing a shirt reading 'MY EX IS MY BIGGEST FAN.' The post spread widely, was liked hundreds of thousands of times, and was covered as genuine by outlets including OK! and People, despite Goth's representatives saying they had repeatedly asked Instagram to remove the account.

Wikimedia bans AI-written text on Wikipedia after months of cleanup efforts failed

Wikipedia editors, who once hoped AI would merely automate formatting and citations, found generative models flooding the encyclopedia with fabricated sources and generic filler text. A Princeton study found roughly 5% of nearly 3,000 new English Wikipedia articles in August 2024 contained substantial AI-generated content. By March 2026, Wikimedia had banned editors from writing articles with large language models, limiting AI use to translation and minor proofreading.

University of Maryland team builds IdeaLens to flag AI-generated research ideas

Researchers led by Rishanth Rajendhran at the University of Maryland have built a tool called IdeaLens that distinguishes human- from AI-generated ideas by analyzing a text's stripped-down outline rather than its wording or style. In tests described in a preprint, the detector correctly identified AI-derived ideas 91% of the time by spotting characteristic argument structures and flow.

Nikon strips AI-generated microscopy video of Small World in Motion win

Nikon revoked the top prize in its 16th annual Small World in Motion contest after scientists flagged Dr. Ning Xu's time-lapse of cilia in a child's airway as likely AI-generated. Researchers pointed to cellular structures that appeared and vanished unnaturally, and a PhD student found what appeared to be an AI-generation watermark on the source image. Xu acknowledged using AI during post-processing to enhance grayscale images but denied generating the cilia or their motion artificially.

Nikon strips Ning Xu's cilia video of Small World in Motion top prize over AI use

Nikon Instruments disqualified the winning entry in its Small World in Motion microscopy competition after determining the video violated rules barring AI-generated content. The video, submitted by National University of Singapore engineer Ning Xu and depicting beating airway cilia, had won first place on 15 September before scientists flagged biologically implausible features on social media. Xu said an AI model was used only to enhance visual presentation below the cilia, not to make anatomical claims.

New term 'meat proxy' describes workers who relay AI output unreviewed

A phrase coined by software engineer Niklas Gruhn in August 2026, 'meat proxy,' describes someone who passes along AI-generated content to others without reading, reviewing, or thinking it through first. The term has begun circulating as shorthand for a growing concern about professionals outsourcing judgment entirely to AI tools rather than using them as a starting point for their own analysis.

OpenAI releases nearly 400 AI-generated math results across 700+ manuscripts

OpenAI published a massive collection of almost 400 mathematical results spanning combinatorics, geometry, number theory, topology, probability and other fields, contained in over 700 manuscripts. The release was so large that OpenAI issued guidance to help researchers navigate its GitHub repository, and some proofs include Lean formalizations for computational verification.

OpenAI publishes 700+ AI-generated math proofs, drawing mathematician backlash

OpenAI released hundreds of papers containing AI-generated mathematical proofs, following earlier controversy over its models' handling of the Navier-Stokes equations. The Association for Human Mathematics issued a statement condemning the release, which was shared and endorsed by Fields Medalist Terence Tao, arguing the dump of files demonstrates power rather than scholarship. OpenAI's announcement cited consultations with an independent group, the Advisory Group on Mathematics and Artificial Intelligence, but did not mention that the group had actually opposed continued testing of its proprietary models on advanced problems.

New site Fleeting generates fake meeting audio and calendar invites

A website called Fleeting lets users generate realistic, AI-voiced meeting recordings and fake calendar invites to simulate being busy at work. The browser-based tool creates roughly 12-minute synthetic meetings with invented participants, AI-generated or stock video faces, and options for looping or back-to-back fake meetings. Calendar invites can be added to Outlook or downloaded, and the site says nothing is sent to its servers.