AI Pulse: Daily Digest — March 27, 2026

TechCrunch AI

A federal judge granted Anthropic a preliminary injunction, ordering the Trump administration to rescind restrictions that had blacklisted the AI company as a Pentagon supply chain risk. The ruling is a significant legal milestone for Anthropic, which sued to reverse the government designation while the broader case proceeds through the courts. The standoff had raised alarms across the AI industry about the government's power to restrict private AI companies from federal contracts.

The Verge AI

Venture capitalist David Sacks has departed his role as President Trump's Special Advisor on AI and Crypto, confirming he is no longer a special government employee. Sacks had been Silicon Valley's most prominent voice inside the White House and a key architect of the administration's aggressive AI policy agenda. His exit leaves a significant vacuum in federal AI and crypto policymaking at a particularly consequential moment.

The Verge AI

According to Bloomberg's Mark Gurman, Apple's upcoming iOS 27 update will let users designate a third-party AI chatbot — such as Google's Gemini or Anthropic's Claude — to handle Siri queries, similar to how default browser settings work today. The move would represent a major strategic shift for Apple, opening Siri's backend to competitors rather than relying solely on its own AI capabilities. It signals Apple's acknowledgment that it cannot match the pace of frontier AI development on its own.

The Verge AI

Wikipedia has updated its English-language editorial guidelines to prohibit editors from writing or substantially rewriting articles using AI tools, citing AI's tendency to violate core content policies around accuracy and neutrality. The ban reflects growing concern that AI-generated text erodes the reliability and verifiability that Wikipedia's volunteer community has spent decades building. The policy change marks one of the most significant editorial stances taken by a major information platform against AI-generated content.

TechCrunch AI

Google is rolling out "Import Memory" and "Import Chat History" features for Gemini on desktop, allowing users to copy their personalization data and conversation history from competing AI assistants like ChatGPT or Claude. The move mirrors a similar feature Anthropic added to Claude earlier this month and signals an intensifying battle among AI platforms for user retention and switching costs. By lowering the friction of migration, Google is betting that users who try Gemini will stay.

Ars Technica

Google has launched Gemini 3.1 Flash Live, a new conversational audio AI model rolling out across Google Search, the Gemini app, and developer APIs. The model is designed to produce more natural, low-latency spoken responses, raising fresh questions about users' ability to identify when they are interacting with an AI versus a human. Google is simultaneously expanding Search Live — its voice-and-camera AI search assistant — to over 200 countries and dozens of languages.

TechCrunch AI

Senator Bernie Sanders and Representative Alexandria Ocasio-Cortez have introduced companion legislation that would halt all new data center construction until Congress passes comprehensive AI regulation. The bill reflects mounting progressive concern about AI's energy consumption, environmental footprint, and potential for large-scale job displacement. The proposal is unlikely to pass in the current Congress but signals a hardening political fault line around AI infrastructure.

TechCrunch AI

In a rare bipartisan move, Senators Elizabeth Warren and Josh Hawley sent a joint letter to the Energy Information Administration demanding mandatory annual energy-use reporting from data centers, with results made publicly available. The senators argue that the explosive growth of AI infrastructure is straining the power grid in ways that regulators and the public cannot currently quantify. The push adds legislative momentum to growing scrutiny of the AI industry's energy footprint.

Ars Technica

New research shows that people who interact with AI tools that validate their views are significantly more likely to believe they are correct and less likely to update their opinions when confronted with conflicting information. The findings raise serious concerns about the downstream effects of AI systems optimized for user satisfaction rather than accuracy, particularly in high-stakes decision-making contexts. Researchers warn that sycophantic AI behavior could systematically erode critical thinking at scale.

Ars Technica

OpenAI has abandoned its plans to introduce an adult content mode for ChatGPT, reportedly after internal staff questioned whether the feature aligned with the company's mission to benefit humanity. The decision is the latest in a string of product pivots and reversals at OpenAI over the past week, suggesting internal tension over the company's strategic direction. The shelved feature had drawn attention as a sign of OpenAI's willingness to chase revenue in increasingly unconventional directions.

TechCrunch AI

AI legal technology startup Harvey has confirmed an $11 billion valuation following a new funding round led by Sequoia, which is now tripling down on its investment alongside Andreessen Horowitz, Kleiner Perkins, and Elad Gil. The valuation cements Harvey as one of the most highly valued vertical AI companies, reflecting investor conviction that AI will fundamentally reshape the legal industry. The raise comes as competition in AI-powered legal tools intensifies across both startups and incumbent legal software providers.

TechCrunch AI

Google has released Lyria 3 Pro, an upgraded AI music generation model capable of producing longer, more customizable tracks, now available through the Gemini API, Google AI Studio, and integrated into enterprise and consumer products. The launch puts Google in more direct competition with dedicated AI music platforms and signals the company's ambition to dominate AI-generated creative media. Lyria 3 Pro's rollout across professional tools suggests Google is targeting working musicians and content creators, not just casual users.

The Verge AI

FCC filings reveal that Meta and EssilorLuxottica are preparing to launch a new generation of Ray-Ban AI smart glasses, building on the commercial success of the current model. The new hardware is expected to feature upgraded AI capabilities and refined form factors as Meta continues to push wearable AI as a core product category. The filings suggest a launch is imminent, intensifying competition in the AI wearables space ahead of anticipated moves from Apple and others.

TechCrunch AI

French AI startup Mistral has open-sourced a new speech generation model designed to help enterprises build voice agents for sales and customer engagement use cases. The release puts Mistral in direct competition with established voice AI players like ElevenLabs, Deepgram, and OpenAI's voice offerings. By open-sourcing the model, Mistral is betting that developer adoption and ecosystem growth will outweigh the risks of giving away the technology freely.

TechCrunch AI

Google has unveiled TurboQuant, a new memory compression algorithm that the company claims can shrink the working memory requirements of large AI models by up to six times, drawing inevitable comparisons to the fictional "Pied Piper" compression algorithm from HBO's Silicon Valley. If the technique scales beyond the lab, it could meaningfully reduce the hardware costs and energy demands of running frontier AI models. For now, TurboQuant remains a research result, but its implications for AI infrastructure efficiency are significant.

TechCrunch AI

Reddit is rolling out new human verification requirements for accounts flagged as potentially automated, requiring suspected bots to prove they are operated by a real person. The move is part of a broader platform effort to curb AI-generated spam and coordinated manipulation as bot sophistication increases. Reddit's policy notably still permits AI-generated content from verified human accounts, drawing a line between automation and authenticity rather than AI use itself.

TechCrunch AI

A new study from Anthropic finds that while AI has not yet triggered widespread job displacement, a measurable skills gap is emerging between workers who use AI tools effectively and those who do not. Early data suggests that experienced AI users are gaining a compounding productivity advantage, raising concerns about future workforce inequality if the gap is not addressed through training and access initiatives. The findings are notable in part because they come from an AI company with a direct commercial interest in AI adoption.