AI persuasion in elections, Anthropic's surge, critical React flaw, digital twin ethics...

MIT Technology Review, Scientific American, and sciencenews.org report that large language model (LLM) chatbots can sway voters’ political views far more effectively than traditional political ads, with new peer-reviewed research showing shifts of up to 25 percentage points in some cases. These studies, published in Nature and Science, found AI chatbots—optimized for persuasion and equipped with fact-based arguments—moved voters toward opposing candidates at unprecedented rates in the US, UK, Canada, and Poland. However, the most persuasive chatbots were also the most prone to spreading misinformation, raising urgent questions about democracy's resilience and the need for robust regulation as 2028 US elections approach. While the EU’s AI Act classifies election-related persuasion as high risk, US legislation remains piecemeal and largely unenforced. Experts warn that both state actors and campaigns will soon have the tools to run automated, individualized persuasion campaigns at scale and low cost, with significant implications for political discourse and electoral integrity. MIT Technology Review MIT Technology Review sciencenews.org Scientific American

The Information analyzes Anthropic's rapid ascent, noting CEO Dario Amodei's disciplined approach and projected profitability by 2028, contrasting sharply with OpenAI’s more aggressive, risk-prone expansion under Sam Altman. Amodei’s thoughtful risk management, showcased at The New York Times DealBook summit, is winning industry confidence as Anthropic secures enterprise-focused deals and maintains cost discipline. TechCrunch reinforces this narrative, highlighting Amodei’s skepticism of “YOLO-ing” strategies among competitors and his belief in sustainable AI economics, while Wired features Anthropic president Daniela Amodei arguing that prioritizing AI safety will ultimately win market trust, despite political pushback against regulation. The Information TechCrunch Wired

A “perfect” security vulnerability with near-100% exploit probability has been uncovered in the popular React JS framework, TechSpot reports. The flaw is exceptionally dangerous due to its ease of exploitation and the potential to compromise both public web servers and private cloud infrastructure. Security experts are advising urgent mitigation efforts, as the widespread use of React in modern web development could magnify the impact of any attack that leverages this vulnerability. TechSpot

Computer World explores the rapid adoption of digital twin technology—from personalized medicine to workforce automation and influencer engagement—enabled by advances in AI and large language models. While digital twins offer benefits like individualized healthcare and efficient research, public trust remains low: a recent survey found nearly 70% of US shoppers would lose trust in brands replacing real feedback with synthetic personas. The backlash is evident as companies like Meta retreat from public-facing AI characters amid user discomfort. Researchers also caution that digital twins currently replicate human responses with only 75% accuracy—no better than generic demographic models—underscoring the need for ethical safeguards and transparency as adoption spreads. Computer World

The Information reports that Fluidstack, a cloud provider supporting Google's AI chip development, is seeking over $700 million in new funding, led by a fund with stakes in CoreWeave and Anthropic. This underscores the escalating race to build AI infrastructure and the growing influence of independent cloud players in enabling major tech companies’ chip and compute ambitions. The Information

The Information reveals that Microsoft is strategizing to maintain software subscription revenues as AI automates more white-collar jobs. The company proposes selling seat-based licenses not just to humans but also to AI agents operating within client organizations, potentially redefining how enterprise software is monetized in the era of agentic automation. The Information

Zdnet and The Register detail OpenAI’s latest alignment experiment, where its GPT-5 Thinking model was rewarded for truthfully “confessing” to lies, hallucinations, or mistakes after generating initial responses. While not prompted by a major incident, the study aims to explore new methods for making future AI systems more transparent and accountable, particularly as questions about AI reliability and liability intensify. Zdnet The Register

At AWS re:Invent, The Register reports, Amazon introduced the Graviton5, its most powerful homegrown CPU yet, packing 192 cores into a single socket. Graviton chips now represent half of all new CPUs added to AWS over the past three years, highlighting Amazon’s accelerating push to optimize its cloud infrastructure with custom silicon and reduce reliance on Intel and AMD. The Register

Tech Radar highlights Nvidia CEO Jensen Huang’s successful lobbying efforts against proposed US export restrictions on AI chips and his warnings about the dangers of state-level AI regulation fragmentation. Meanwhile, Nvidia has ended Linux “Game Ready” driver support for its aging GTX 900 and 10-series GPUs, though quarterly security patches will continue, signaling a shift of focus to newer hardware generations. Tech Radar Tech Radar

Wired reports that Cloudflare has blocked 416 billion AI bot requests since July, reflecting mounting tensions between internet infrastructure companies and AI developers scraping web content for model training. CEO Matthew Prince claims these efforts are producing tangible results, as publishers and platforms seek to protect their data from unauthorized AI ingestion. Wired

The Scientist covers new research from the University of Bristol indicating that key eukaryotic traits—foundational for complex life—may have emerged nearly a billion years earlier than previously believed. By analyzing gene duplication events across the tree of life, scientists suggest that the blueprint for eukaryotic cells predates the arrival of mitochondria, reshaping our understanding of early evolution and the timeline of life on Earth. The Scientist