All articles tagged with #ai detection
Anthropic detailed its Claude watermarking system, rooted in a secret-key token approach inspired by Google’s SynthID, claiming negligible speed impact and that watermarking could appear even when Claude edits text; the watermark is intended to help comply with the EU AI Act. Despite assurances, some Claude users on X report canceling subscriptions in protest, though Anthropic says overall cancelations have not spiked.
The administration paused more than $1 billion in Medicaid funding to California and Minnesota over suspected fraud and noncompliance, using AI and advanced analytics to flag suspicious payments; California faces roughly $867 million and Minnesota about $200 million in the freeze, following an earlier $243 million pause in Minnesota. The move comes as the HHS expands authority to exclude bad actors from federal programs and states face pressure to document claims, with Minnesota having previously sued over withholding funds.
Meta's Muse Image introduces an invisible Content Seal watermark and a detector, but Reuters found the detector correctly flags full-size Muse Image images yet drops to about 55% accuracy after cropping, highlighting a gap between generation and detection. Privacy concerns over using public-profile photos led Meta to remove that feature, as the company pursues future tools like Muse Video.
An AI-assisted analysis of seafloor bathymetry led by Andrea Verolino identifies 73 previously unknown submarine calderas, potentially tripling the world's known count and revealing counts at mid-ocean ridges, volcanic arcs, and interior seafloor. The study provides a transparent framework for mapping submarine volcanism and assessing hazards, though it does not confirm whether any of these calderas are currently active.
A Yale study finds oleic acid from olive oil can fuel pancreatic ductal adenocarcinoma via ferroptosis, while polyunsaturated fats may slow disease; separately, Mayo Clinic's REDMOD AI can detect pre-diagnostic pancreatic cancer on routine CTs with about 73% sensitivity; CNIO reports a triple-drug combo blocking RAF1, EGFR, and STAT3 that eliminates tumors in experimental models, highlighting advances in precision nutrition, early detection, and combination therapies against this deadly cancer.
Three of the five regional winners of the 2026 Commonwealth Short Story Prize are facing allegations of using AI to craft their stories, with Pangram flagging Jamir Nazir’s The Serpent in the Grove as AI-generated and other entries as AI-generated or AI-assisted. Granta and the Commonwealth Foundation defend the integrity of the judging, note trust in unpublished fiction, and have added disclaimers to the winning stories; AI-detection tools are deemed imperfect, highlighting broader tensions over AI’s role in creative writing, including a judge accused of AI-assisted marketing copy.
Deezer reports AI-generated music has surged to about 75,000 tracks per day (over 2 million per month), making up 44% of all new uploads, while AI streams remain only 1-3% of total plays and about 85% of those are fraudulent and demonetized; Deezer has stopped storing hi‑res AI tracks, expanded platform-wide AI detection (and licensed the tech to others), and says it can identify 100% of AI content from leading models, having tagged over 13.4 million AI tracks in 2025, as industry debates how to protect artists’ rights and ensure transparency, with peers like Apple Music and Spotify pursuing disclosure standards rather than detection.
Deezer reports about 75,000 daily AI-generated song submissions, roughly 44 percent of all uploads, yet AI tracks drive only 1–3 percent of streams as the service demonetizes AI music, labels AI-made tracks, and calls for industry transparency while rivals pursue their own AI policies.
New advanced AI-generated images of receipts are increasingly being used to commit expense fraud, prompting companies to adopt AI-based detection methods that analyze image metadata and contextual data to identify fakes, as the technology becomes more convincing and accessible.
YouTube is rolling out a new AI 'likeness detection' tool for creators in its Partner Program to identify and report unauthorized uploads using their likeness, including deepfakes, with the feature currently in early access and expanding over the coming months.
A new report indicates that over 50% of recent internet articles are now AI-generated, with the share plateauing around this level since late 2024, suggesting a stabilization in AI content production and potential shifts in content creation practices.
AI-generated articles briefly outnumbered human-written ones online but are now roughly equal, with AI content still comprising a minority of search rankings and user trust remaining low. Researchers highlight the difficulty in distinguishing AI from human content and note that humans prefer human-written material, though AI's role in content creation continues to grow.
Deezer reports that nearly 28% of daily uploaded tracks are fully AI-generated, with over 30,000 such tracks received daily, though they constitute only 0.5% of streams. The platform is actively filtering out AI content from recommendations and playlists to minimize its impact, amid concerns about fraudulent activity and copyright issues. Deezer's efforts include AI detection tools and transparent tagging, as the industry grapples with the rise of generative AI in music.
A team of computer scientists developed an AI-based classifier to identify questionable open access scientific journals, which often prioritize fees over editorial quality, aiming to combat the proliferation of predatory publishing and improve scientific integrity.
A new AI framework called CTCAIT can detect neurological disorders like Parkinson's, Huntington's, and Wilson disease with over 90% accuracy by analyzing speech patterns, offering a promising non-invasive tool for early diagnosis and monitoring across multiple languages.