AI Reporter - MARCH 2026

MARCH 2026PAGE 2AI UpdatePAGE 3AI in BusinessNapster relaunches as AI musicplatformNew AI model from MGB could predictdementia risk and moreAmazon’s AI push aims to streamlineHollywoodStudy finds AI stethoscope doublesdetection of serious valve diseasePAGE 4MSD and Mayo Clinic partner onAI-powered drug discoveryYouTube brings conversational AI tosmart TVs and streaming devicesPAGE 5AI Litigation & RegulationLITIGATIONCompany claims Google’s “Gemini”expansion violates trademark in FloridalawsuitLawsuit filed in Northern District ofCalifornia alleges Nvidia illegally scrapedYouTube videos for AI trainingNew York federal court overturns orderrequiring OpenAI to disclose attorneycommunicationsPAGE 6Engrain accuses Beans.ai of copyingmaps to launch rival product inColorado lawsuitNorthern California court blocks OpenAIfrom using “cameo” in Sora videofeatureLawsuit filed in California federal courtchallenges AI training practices atRunway AICalifornia federal judge dismisses xAI’strade secret claimsPAGE 7REGULATIONADLM warns of AI bias risksCLEAR Act seeks greater transparencyin AI trainingNIST launches initiative to advancesecure interoperable standards forautonomous AI agentsIn This IssueA Publication of Benesch’sAI CommissionAI Reporterbeneschlaw.comThe content of the Benesch, Friedlander, Coplan & Aronoff LLP AI Reporter is for general information purposes only. It does not constitutelegal advice or create an attorney-client relationship. Any use of this newsletter is for personal use only. All other uses are prohibited.©2026 Benesch, Friedlander, Coplan & Aronoff LLP. All rights reserved.beneschlaw.com 2New lawsuits accuse major AI developers of using copyrighted content to train their models,including claims that Nvidia secretly scraped millions of protected YouTube videos to build its Cosmosvideo model and Runway AI purportedly bypassed YouTube’s download protections to gathertraining data for its generative systems. In these California federal courts cases, plaintiffs argue thecompanies ignored copyright rules and platform terms of service and commercially benefited from theunauthorized data. They now face proposed class actions seeking damages, dataset destruction andrepresentation of affected creators and authors across the United States.Meanwhile, recent court decisions delivered mixed outcomes for OpenAI. In a federal New Yorkcopyright case, a federal judge blocked an order requiring the company to disclose internal attorneycommunications about deleted training datasets, ruling that OpenAI had not waived the attorney-clientprivilege and its explanations for deleting Books1 and Books2 did not put its legal advice “at issue.” Ina separate case from the Northern District of California, another judge issued a preliminary injunctionbarring OpenAI from using the term “cameo” for a Sora video-generation feature, finding that Cameo islikely to prevail on its trademark claim given evidence of consumer confusion and potential reputationalharm linked to Sora-generated deepfakes. Meanwhile, also in the Northern District of California, xAI’slawsuit alleging that OpenAI induced employees to steal trade secrets was dismissed, with the judgeconcluding that the claims lacked detail and did not show that any confidential xAI materials were usedat OpenAI or that the company encouraged wrongdoing.On the regulatory side, the Association for Diagnostics & Laboratory Medicine (ADLM) warnedthat AI systems used in laboratory medicine risk reinforcing health inequities if trained on biased orinconsistent data, urging federal regulators to update oversight to ensure safety, effectiveness andfairness. At the same time, the CLEAR Act, a bipartisan Senate bill, seeks greater transparencyin GenAI training by requiring companies to publicly disclose copyrighted materials used in modeldatasets, with fines for noncompliance. Elsewhere, the National Institute of Standards and Technology(NIST) launched an initiative to promote secure, interoperable autonomous AI agents by developingindustry-led standards, open-source protocols and research on authentication, aiming to ensuretrustworthy integration of AI agents across digital systems and inviting public feedback through futureguidelines and listening sessions.These and other stories appear below.Steven M. SelnaPartnerAI UpdateMARCH 2026AI Reporterbeneschlaw.com 3Napster relaunches as AI music platformUnder new ownership, Napster relaunched its app to focusentirely on AI-generated music and podcasts, eliminatingits traditional record-label catalog. The platform now offersAI-generated content tailored to every genre, mood andpersonal preference, leveraging real-time creation tools.CEO John Acunto stated that Napster is moving awayfrom relationships with major labels, criticizing them forsuppressing content and data ownership. Napster is involvedin ongoing lawsuits about royalties owed to Sony Music andSoundExchange.Source: MusicallyNew AI model from MGB could predictdementia risk and moreResearchers at Mass General Brigham have developedBrainIAC, a self-supervised AI model trained on over49,000 brain MRI scans. This foundational model uses keyhealth indicators to accurately predict neurological healthoutcomes, such as dementia risk, brain cancer survival, andother diseases, even when training data is limited. BrainIACidentifies key features from MRI datasets and is adaptableto other healthcare applications, potentially improving carein settings where annotated medical data is scarce. Thistechnology demonstrates how AI can enhance patient careand treatment options by extracting insights from diverse andlimited datasets, addressing challenges in access to care andsupporting the development of new diagnostic tools.Source: Healthcare IT NewsAmazon’s AI push aims to streamlineHollywoodAmazon is developing new AI tools aimed at accelerating andstreamlining the production of movies and TV shows. Led byveteran executive Albert Cheng at Amazon MGM Studio, theinitiative seeks to cut costs and enhance the creative processin entertainment. Amazon will launch a closed beta programin March, inviting industry partners to test these AI tools, withresults expected by May. The move comes amid industryconcerns that AI could lead to job losses and fundamentallychange Hollywood’s landscape, raising questions about theimpact of AI on creative roles and intellectual property in theentertainment sector.Source: Reuters (sub. req.)Study finds AI stethoscope doubles detectionof serious valve diseaseA recent prospective study published in the EuropeanHeart Journal Digital Health found that an AI-enableddigital stethoscope significantly improved the detection ofmoderate-to-severe valvular heart disease (VHD) by primarycare providers compared to standard stethoscopes. The AIsystem achieved a sensitivity of 92.3% for detecting audibleVHD, versus 46.2% for standard care, and identified twice asmany previously undiagnosed cases. While specificity wasslightly lower, the findings suggest AI-assisted auscultationcould serve as a valuable screening adjunct, potentiallyenabling earlier intervention and improved patient outcomes.Source: News Medical Life SciencesSydney E. AllenSenior Managing AssociateMARCH 2026AI ReporterAI in Businessbeneschlaw.com 4MSD and Mayo Clinic partner on AI-powereddrug discoveryThe collaboration grants MSD (known as Merck & Co in theU.S. and Canada) access to the Mayo Clinic Platform, whichincludes deidentified laboratory results, medical imaging,clinical notes, molecular data, registries and biorepositories,along with advanced AI and analytics tools. The partnershipaims to enhance disease understanding, improve targetidentification and accelerate early development decisions byintegrating high-quality clinical data and AI-enabled insights.The alliance is expected to speed up the development ofnew therapies and improve patient outcomes, while alsohighlighting the importance of data privacy and security in theuse of AI for healthcare and life sciences applications.Source: Pharma PhorumYouTube brings conversational AI to smart TVsand streaming devicesYouTube expanded its conversational AI feature to smartTVs, gaming consoles and streaming devices, allowingusers to interact with AI directly on their television screens.The tool allows viewers to ask questions about the contentthey are watching, such as details about a song or recipe,without pausing or leaving the app. The AI assistant providessuggested questions and instant answers, supportingEnglish, Hindi, Spanish, Portuguese and Korean.Source: TechCrunchMARCH 2026AI ReporterAI in Business (cont’d)beneschlaw.com 5LITIGATIONCompany claims Google’s “Gemini” expansionviolates trademark in Florida lawsuitP2B Trading sued Google in the Middle District of Florida,alleging the tech giant’s expanding use of the “Gemini”name will destroy P2B’s brand and violate its decades-oldtrademark. P2B, a longtime seller of Gemini-branded audioequipment, has held rights to the Gemini mark for speakersand other audio products since 1998 and says Google’s planto release a Gemini-branded home speaker in 2026 directlyconflicts with its core market. The company argues Google’smassive market power, combined with its rebranding andintegration of Gemini into hardware like earbuds, will createwidespread consumer confusion. P2B also claims Googlepreviously assured it would not use the name on audioequipment and later reversed course. The lawsuit seeksto stop Google from using the Gemini name, cancel itstrademarks, and recover profits and damages.Source: Law 360 (sub. req.)Lawsuit filed in Northern District of Californiaalleges Nvidia illegally scraped YouTube videosfor AI trainingA new lawsuit claims that Nvidia covertly harvested millionsof YouTube videos to train its foundational AI video model,Cosmos. The plaintiff podcaster alleges the companybypassed YouTube’s technical protections and terms ofservice, which prohibit scraping and bulk downloading.The suit also says Nvidia engineers discussed large-scaledata gathering in internal Slack channels and developedworkarounds to extract video “clips” without creator consent.These datasets reportedly formed the basis for Cosmosand supported other Nvidia products, giving the companystrong commercial incentives to acquire massive amounts ofcontent. The podcaster seeks to represent all U.S. YouTubecreators whose videos were allegedly scraped.Source: Law 360 (sub. req.)New York federal court overturns orderrequiring OpenAI to disclose attorneycommunicationsA New York federal judge blocked an order that would haverequired OpenAI to hand over internal communications withits in-house attorneys about deleted training datasets. Thecourt ruled that OpenAI did not waive attorney-client privilege,rejecting findings that the company’s statements aboutdeleting datasets “due to non-use” revealed legal advice. Thejudge also disagreed that OpenAI’s denial of willful copyrightinfringement placed its state of mind “at issue,” noting thatsimply denying allegations is not the same as using legaladvice to assert good-faith conduct. The decision reverses aprior ruling that OpenAI must disclose 2022 communicationsabout the deletion of datasets and references to the shadowlibrary LibGen. The broader case involves claims that OpenAIand Microsoft trained AI models on copyrighted workswithout permission, but the judge concluded sanctions foralleged shifting positions on privilege were unwarranted.Source: Law 360 (sub. req.)Carlo LipsonAssociateMARCH 2026AI ReporterAI Litigation & Regulationbeneschlaw.com 6Engrain accuses Beans.ai of copying maps tolaunch rival product in Colorado lawsuitA mapping technology company claims a former potentialpartner replicated thousands of its property maps to fasttrack a competing product. Engrain alleges that Beans.aileveraged access granted during a brief partnership effort toreverse engineer its detailed unit maps and develop the freeWebWidget tool, despite anti-copying restrictions. Accordingto the complaint filed in the district court of Colorado, Beansaggressively marketed WebWidget, even sending promotionsto Engrain itself in 2025. Engrain also asserts a major industrypartner began adopting Beans’ free immersive maps,prompting several customer cancellations. The companypoints to identical layout quirks and duplicated GPS errors asevidence that Beans mimicked its flagship SightMap product.Beans.ai’s co-founder denies wrongdoing, calling the lawsuita reaction to his company’s recent growth and emphasizingthat Beans uses AI—not manual processes—to generatemaps. The suit includes claims of copyright infringement,breach of contract, unjust enrichment and vicariousinfringement.Source: Law 360 (sub. req.)Northern California court blocks OpenAI fromusing “cameo” in Sora video featureA Northern California federal judge issued a preliminaryinjunction preventing OpenAI from using the term “cameo”to describe a feature in its Sora video generator. The courtdetermined that the usage threatens the reputation andgoodwill of the celebrity video platform Cameo. OpenAIargued the term was descriptive, but the court rejectedthat defense, instead citing the strength of Cameo’s brand,extensive media presence, and evidence of actual consumerconfusion, including misdirected messages, incorrect socialmedia tagging and assumptions that the two companieswere partnered. The judge also noted that Sora generateddeepfakes of public figures could harm Cameo’s associationwith authentic celebrity interactions.Source: Law 360 (sub. req.)Lawsuit filed in California federal courtchallenges AI training practices at Runway AIA YouTube creator has filed a proposed class action allegingthat Runway AI unlawfully accessed and downloadedYouTube videos to train its GenAI models. The complaintclaims the company bypassed platform protections thatprevent downloading, violating YouTube’s terms of serviceand ignoring copyright restrictions. Instead of seekingpermission or paying creators, Runway AI allegedly scrapedvideo files to improve its commercial AI products. The CentralDistrict of California suit brought claims under the DigitalMillennium Copyright Act, California’s Unfair Competition Lawand unjust enrichment. This case adds to a growing wave oflawsuits accusing AI developers of using copyrighted materialwithout authorization to train their systems.Source: Law 360 (sub. req.)California federal judge dismisses xAI’s tradesecret claimsA Northern District of California federal judge threw out xAI’slawsuit accusing OpenAI of inducing employees to stealtrade secrets. The judge noted that while several departingemployees kept xAI source code or confidential materials onpersonal devices, the complaint’s claims were too vague andunsupported to assert they used that information at OpenAIor that OpenAI encouraged any theft. Only one accusedemployee joined OpenAI, and another’s attempted accessto xAI documents was blocked. Other former staffers werenot alleged to have taken or used confidential data at all.The judge concluded that xAI failed to show misconduct byOpenAI itself, despite the timing of multiple departures andcommunications with a recruiter.Source: Law 360 (sub. req.)MARCH 2026AI ReporterAI Litigation & Regulation (cont’d)beneschlaw.com 7REGULATIONADLM warns of AI bias risksThe Association for Diagnostics & Laboratory Medicineissued a statement highlighting the risks associatedwith the use of AI in laboratory medicine, particularlyfor historically marginalized patient populations. Thestatement urges Congress and federal agencies to updatelaboratory regulations and implement policies to ensure AIclinical systems are both safe and effective. While AI hasthe potential to enhance diagnostic accuracy, improvelaboratory efficiency, and enable more precise datadriven clinical decision-making, concerns remain about AImodels perpetuating societal biases if trained on limited orinconsistent data. These biases can lead to misclassificationof diseases, health inequities and underestimation of riskin marginalized groups, thus underscoring the need forresponsible oversight and consistent data standards.Source: PR NewswireCLEAR Act seeks greater transparency in AItrainingA bipartisan Senate proposal would require technologycompanies to publicly disclose the copyrighted materialsthey use to train GenAI models. The CLEAR Act, introducedby Senators Adam Schiff and John Curtis, aims to protectcreators while allowing innovation to continue responsibly.The bill mandates companies to submit detailed summariesof copyrighted works included in training datasets to the U.S.Copyright Office, which would maintain a public databaseof these disclosures. Noncompliance would trigger fines ofat least $5,000 per undisclosed instance. Creative industrygroups, including musicians’, writers’ and artists’ unions,strongly support the measure as a path toward greatertransparency and fair compensation.Source: Law 360 (sub. req.)NIST launches initiative to advance secureinteroperable standards for autonomous AIagentsAs AI increasingly handles tasks such as coding, emailmanagement and online transactions, NIST launcheda new initiative to support the secure and interoperabledevelopment of autonomous AI agents. The programfocuses on preventing fragmentation by promoting consistenttechnical guidelines and smooth integration with existinginfrastructure for AI agents to interact reliably across digitalsystems. This collaboration with federal partners and theNational Science Foundation will guide efforts acrossthree pillars: global leadership in standards, communitydriven protocol development and research into secureauthentication and authorization. NIST plans to releaseadditional guidelines and solicit public feedback throughRFIs, listening sessions and sector-specific discussions toencourage confident adoption across industries.Source: NISTMARCH 2026AI ReporterAI Litigation & Regulation (cont’d)Are you interested in a particular topic that you wouldlike to see covered in the Reporter? If so, please letus know.Steven M. [email protected]: 628.600.2261Sydney E. AllenSenior Managing [email protected]: 628.600.2229Carlo [email protected]: 628.600.2247