In the Loop: Week Ending 9/13/25

Last Week in AI: Adoption Slowdown, Brain Drain, AI Critterz

Corporate AI Adoption Shows First Major Decline Since 2023

U.S. Census Bureau data reveals AI usage among large companies is dropping for the first time since tracking began, with adoption at firms over 250 employees falling from nearly 14% in mid-June to under 12% in August. The decline represents the largest drop-off since the survey launched in November 2023, surveying over 1.2 million firms nationwide. The downturn comes despite continued AI hype driving tech stocks to record highs. A staggering 95% of companies using AI report the technology has failed to generate any new revenue, raising questions about enterprise AI's viability. The summer brought additional disappointments, including GPT-5's underwhelming performance on benchmarks and companies scrambling to rehire workers previously laid off for AI automation. Unless enterprise AI demonstrates clear value, this decline could signal broader disillusionment with corporate AI adoption.

U.S. Student Test Scores Hit Historic Lows as AI Era Begins

American high school seniors' math and reading scores dropped to their lowest levels in over 20 years, according to the latest National Assessment of Educational Progress results. The 2024 scores show 32% of 12th graders scored below "basic" in reading – meaning they couldn't find details in text to understand its meaning – while 45% scored below basic in math, the highest percentage since 2005. Only 33% of seniors were considered academically prepared for college-level math, down from 37% in 2019. While experts warn against drawing definitive conclusions due to multiple factors including COVID-19's lingering effects, the timing coincides with rapid AI adoption in schools. Educators point to potential underlying factors including increased screen time, shortened attention spans, and decline in reading longer-form writing, raising questions about AI's educational impact.

Judge Rejects Record $1.5 Billion Anthropic Copyright Settlement

Federal Judge William Alsup rejected Anthropic's proposed $1.5 billion settlement with authors who accused the company of using pirated books to train its Claude chatbot, calling the agreement "nowhere close to complete." The settlement, which would have been the largest copyright recovery in history, proposed paying authors roughly $3,000 for each of the estimated 465,000 books covered. Alsup expressed concerns that the deal could be forced "down the throat of authors" without proper input and demanded clearer details about works lists, notification processes, and claim forms. The judge set September 15 and October 10 deadlines for final works lists and document approval. Anthropic had wrongfully acquired 7 million digitized works from piracy websites like Library Genesis, though training AI on legally acquired books wasn't ruled illegal.

Dictionary Wars: Britannica and Merriam-Webster Sue Perplexity

Encyclopedia Britannica and Merriam-Webster filed a federal lawsuit against Perplexity AI, accusing the startup of copyright and trademark infringement through its "answer engine" that systematically copies their content without permission. The Chicago-based publishers claim Perplexity's system "free rides" on their investments by cannibalizing traffic to their sites while providing users with potentially inaccurate AI-generated summaries. The lawsuit includes examples showing how Perplexity searches on topics like druids, quantum physics, and Amelia Earhart relied heavily on Britannica information. This adds to mounting legal pressure on Perplexity, which already faces similar lawsuits from News Corp's Dow Jones and the New York Post, plus Japanese media companies. The timing follows Anthropic's recent copyright settlement, potentially signaling that publishers see AI litigation as increasingly winnable.

Hollywood AI Revolution: From Critterz to Factory Lines

AI is rapidly transforming movie production as OpenAI backs "Critterz," a feature-length animated film targeting a 2026 Cannes debut. The project promises to cut typical three-year animation timelines to just nine months with a budget under $30 million – far less than Pixar's $200 million productions. Using tools like DALL-E and Sora, the film follows forest creatures disrupted by a mysterious stranger. Meanwhile, the broader "AI movie factory" is ramping up across Hollywood. Amazon-backed startup Showrunner plans to restore 43 minutes of lost footage from Orson Welles' "The Magnificent Ambersons," though Welles' estate objects to the "purely mechanical exercise." The Sphere venue deployed AI to reformat "The Wizard of Oz" with an $80 million budget. Directors at Toronto International Film Festival pushed back, emphasizing uniquely human storytelling qualities.

FTC Launches Major Probe Into AI Chatbot Safety for Kids

The Federal Trade Commission issued orders to seven major AI companies – including Google, Meta, OpenAI, Snap, xAI, and Character.AI – demanding information about how their chatbot companions potentially harm children and teenagers. The inquiry focuses on AI systems that "effectively mimic human characteristics, emotions, and intentions" and act as companions, which may prompt young users to trust and form relationships with the technology. The probe comes after mounting controversies linking AI interactions to severe mental health crises, including lawsuits alleging chatbots contributed to teen suicides through "sycophantic" behavior that validated dangerous thoughts. The FTC wants to understand how companies measure safety, limit negative impacts on minors, and inform parents about risks. Companies must respond by late September, with findings potentially shaping future regulations around AI companion technologies.

Senator Cruz Unveils AI Regulatory "Sandbox" to Boost Innovation

Senate Commerce Chairman Ted Cruz introduced legislation creating a federal regulatory "sandbox" that would allow AI companies to apply for temporary waivers from federal regulations that impede innovation. The 41-page bill directs the White House Office of Science and Technology Policy to establish the program, letting participants test AI technologies for up to two years with potential extensions. Companies could identify obstructive regulations and request modifications while maintaining responsibility for mitigating health, safety, and consumer risks. The legislation reflects concerns that excessive regulation could cause the U.S. to fall behind China in AI development. Cruz emphasized that the sandbox "is not a free pass" and companies must still follow existing laws. Federal agencies would have 90 days to approve or deny applications.

California Passes Groundbreaking AI Chatbot Safety Bill

California's legislature passed pioneering legislation targeting AI companion chatbots, making it the first state to legally require mandatory safety protocols for systems designed to fulfill users' social needs. The bill bans chatbots from engaging minors in conversations about suicide, self-harm, or sexually explicit content, while requiring companies to send mandatory alerts every three hours reminding young users they're talking to AI, not humans. Starting January 1, 2026, the legislation establishes unprecedented accountability measures allowing users to sue AI companies for up to $1,000 per violation plus damages and attorney's fees. Major tech companies including Meta, Google, and Amazon oppose the bill, while OpenAI urged Governor Newsom to reject it in favor of weaker federal frameworks. Only Anthropic publicly supported the legislation.

Salesforce CEO Admits AI Eliminated 4,000 Customer Support Jobs

Salesforce CEO Marc Benioff revealed his company cut 4,000 customer service roles – nearly half its support staff – as AI agents stepped in to handle over a million customer conversations. Speaking on The Logan Bartlett Show podcast, Benioff said he reduced headcount from 9,000 to 5,000 because he "needs less heads" now that technology can manage customer interactions and has reduced support costs by 17%. The admission represents a dramatic reversal from his previous public stance that AI wouldn't lead to white-collar job losses. Benioff noted that AI agents are already doing 30-50% of work within Salesforce and that he's examining "every single function" to identify automation opportunities. The cuts contribute to over 64,000 tech layoffs this year, with Benioff now openly acknowledging what other CEOs deny about AI's workforce impact.

Robotics Investment Boom Reaches Decade-High Momentum

After years of struggling to attract venture capital, robotics startups are experiencing unprecedented investor interest, with $6 billion poured into the sector during the first seven months of 2025. Eclipse partner Seth Winterroth traces the current momentum back to Amazon's 2012 acquisition of Kiva Systems, which "launched 1,000 robotic startups" and attracted engineering talent to the sector. The surge isn't just about AI hype—falling hardware costs for sensors, compute, and batteries have made robotics more viable for scaling. VCs are particularly bullish on manufacturing, healthcare, and eldercare applications, while remaining skeptical about consumer humanoids. As Cybernetix Ventures' Fady Saad notes, even successful consumer robot company iRobot "failed to come up with a second act," highlighting the challenges in that market segment.