Meta just confirmed it won't launch its upcoming multimodal Llama model—or future multimodal versions—in the EU, blaming an unpredictable regulatory environment under GDPR. European firms won't get access even under an open license, though a larger text-only Llama 3 model will still release there.
UK antitrust regulators just opened a formal merger investigation into Microsoft’s hiring spree of Inflection AI staff. The watchdog is evaluating whether poaching co-founder Mustafa Suleyman and most of the startup's team functions as an unapproved backdoor acquisition. The CMA set a September 11 deadline to decide on a deeper review, putting Big Tech's clever AI deal structures on notice.
A bipartisan group of US senators just introduced the COPIED Act to protect artists and journalists from AI scrapers. The bill directs NIST to set provenance standards, makes it illegal to tamper with or remove digital watermarks, and lets creators sue platforms that train models on their protected work. Why it matters: If passed, removing AI tags won't just be sneaky—it'll be a federal offense.
OpenAI just teamed up with Los Alamos National Laboratory to test how multimodal AI models perform inside real-world bioscience labs. The evaluation puts voice and vision capabilities to the test on wet-lab tasks like cell culture and genetic transformation, measuring how much AI assists novices and experts alike. Why it matters: Assessing dual-use biological risks in hands-on environments is becoming a critical front in frontier AI safety.
Microsoft just gave up its observer seat on OpenAI's board, and Apple has dropped its own plans to join. The sudden exit comes amid mounting antitrust scrutiny from US and EU regulators over Big Tech's grip on generative AI. Both companies are keeping their technology partnerships intact, just without the paper trail of boardroom oversight.
OpenAI just started blocking API traffic from unsupported regions like China, sending developers scrambling for alternatives. Domestic giants like SenseTime, Baidu, and Zhipu AI are seizing the moment, giving away up to 150 million free tokens and white-glove migration services. While it triggers a rush toward domestic independence, chip shortages remain a serious bottleneck.
A Phoenix police officer just pulled over a Waymo robotaxi for driving into oncoming traffic, only to find the driver's seat completely empty. The officer was unable to issue a citation, writing in dispatch logs that they were "unable to issue citation to computer." Waymo blamed confusing construction signage, spotlighting a bizarre legal loophole for driverless vehicles.
Google researchers just published a paper warning that generative AI is flooding the web with fake images and videos to sway public opinion and run scams. The team found the barrage of synthetic slop is eroding trust in digital information while giving public figures an easy excuse to dismiss real evidence as AI-generated. Why it matters: It is an ironic admission from a company rushing to bake AI into everything.
OpenAI just patched a glaring security flaw in its ChatGPT macOS app, which was previously storing user conversations in unencrypted plain text on local drives. That meant any malicious app with machine access could easily read your chat history. Following external reports, OpenAI quietly pushed an update that finally encrypts local conversation records.
Google just made it easier for political campaigns to disclose AI-generated content. Instead of requiring advertisers to insert their own notices, Google will now automatically slap an in-ad disclosure on YouTube Shorts, feeds, and in-stream video whenever an advertiser checks the "altered or synthetic" box. Other ad formats will still require manual disclaimers.
Kwai's research team just open-sourced LivePortrait, an efficient portrait animation model that brings still human or animal photos to life using driving videos. Unlike diffusion models, it uses implicit keypoints to clock in at just 12.8ms per frame on an RTX 4090.
Cloudflare just rolled out a free, one-click toggle to block all AI bots, scrapers, and crawlers across its entire network. Rather than just relying on the honor system of robots.txt, its machine learning models automatically flag and block bots that spoof legitimate browsers to steal training data. Why it matters: Site owners finally have an effortless way to stop AI labs from harvesting their content for free.
Apple is set to join OpenAI's board as an observer, tapping App Store chief Phil Schiller for the role as part of their partnership deal. While Schiller won't get voting power, he can attend meetings alongside rival Microsoft, which holds a similar seat. That could make for some awkward boardroom moments whenever sensitive deals come up.
Perplexity is facing fierce backlash from Forbes and Wired, with both outlets accusing the AI search startup of covertly ignoring robots.txt rules to scrape content and republishing original scoops verbatim. Perplexity defends its actions as fair use tied to direct user queries, but the tension between web scrapers and original creators is escalating quickly.
Figma just disabled its new Make Design AI feature after users caught it repeatedly spitting out UI that looked virtually identical to Apple's Weather app. CEO Dylan Field denied training on existing apps, blaming the issue on a rushed design system with low variability. The tool will stay offline until Figma completes a full QA pass.
Suno just launched its official iOS app on the Apple App Store, bringing full text-to-music creation to mobile devices. Users can type in text prompts to generate fully produced songs with vocals and instrumentation in minutes. The launch arrives amid intense legal pressure, as major record labels sue the startup over alleged copyright infringement in its training data.
Anthropic just launched a program to fund third-party organizations building next-generation AI evaluation benchmarks. Current tests do a poor job capturing real-world usage, so the Claude creator is offering grants for evaluations targeting advanced risks like cyberattacks, nuclear weapons, and deception, connecting accepted teams directly with its red team and safety experts.
YouTube just updated its privacy policy to allow people to request the removal of AI-generated content that simulates their face or voice. Victims must file a first-party complaint under privacy violations, giving uploaders 48 hours to remove or blur the footage before platform review kicks in. Why it matters: It treats unauthorized likeness cloning directly as a personal privacy breach rather than broad misinformation.
Meta just changed its 'Made with AI' tag to 'AI info' across Instagram, Facebook, and Threads after photographers complained that real photos retouched in Photoshop were being mislabeled as AI creations. Clicking the new tag now clarifies whether an image was fully generated or simply modified, solving a major headache for creators.
Microsoft AI chief Mustafa Suleyman sparked major pushback after arguing on CNBC that content published on the open web has essentially been "freeware" under a social contract since the 90s. Legal experts were quick to dispute the claim, pointing out that US copyright attaches automatically upon creation and isn't waived just because something is online.
