AI Pulse: Daily Digest — April 2, 2026

TechCrunch AI

OpenAI has closed a massive $122 billion funding round led by Amazon, Nvidia, and SoftBank, with $3 billion coming from retail investors — an unusual move for a pre-IPO company. The round values OpenAI at $852 billion, cementing its position as one of the most valuable private companies in history as it edges closer to a public offering.

Ars Technica

A software update to Claude Code accidentally shipped with a source map file exposing over 512,000 lines of TypeScript code, giving the public an unexpected window into Anthropic's product roadmap. Among the revelations: an always-on persistent agent, a stealth "Undercover" mode, and a Tamagotchi-style virtual assistant called Buddy. Anthropic scrambled to issue DMCA takedowns — accidentally sweeping up thousands of unrelated GitHub repositories in the process before retracting most notices.

TechCrunch AI

In attempting to suppress its accidentally leaked Claude Code source, Anthropic issued overly broad DMCA takedown notices that swept up thousands of unrelated GitHub repositories. Executives acknowledged the error and retracted the bulk of the notices, but the episode compounded an already turbulent stretch for the company.

The Verge

Multiple Apollo Go robotaxis operated by Baidu simultaneously stopped moving in Wuhan, China, stranding passengers inside vehicles, blocking highways, and contributing to at least one accident. Local police confirmed receiving numerous reports of the autonomous vehicles freezing mid-route, raising fresh questions about the reliability and safety of large-scale robotaxi deployments.

MIT Tech Review

A growing global workforce of gig workers — including medical students in Nigeria and freelancers across the developing world — are recording themselves performing everyday physical tasks to generate training data for humanoid robots. The work mirrors earlier waves of data-labeling gig labor that powered language models, but now extends into the physical domain as robotics companies race to teach machines how to move and manipulate objects in the real world.

TechCrunch AI

Meta's forthcoming Hyperion AI data center will draw power from ten newly built natural gas plants, a scale of fossil fuel consumption that rivals the electricity needs of an entire U.S. state like South Dakota. The announcement underscores the mounting tension between Big Tech's AI ambitions and its previously stated sustainability commitments.

TechCrunch AI

Cognichip has secured $60 million in funding to build AI systems capable of designing the very chips that power AI workloads — a recursive bet on automating one of the most complex and expensive engineering processes in the industry. The startup claims its approach can slash chip development costs by more than 75% and cut timelines in half, potentially reshaping the semiconductor supply chain.

Ars Technica

Switzerland's finance minister has filed a criminal defamation complaint against xAI's Grok chatbot, alleging that its AI-generated "roast" feature produces vulgar and degrading content targeting women. The lawsuit is one of the first government-level legal actions against a generative AI product's specific feature, and arrives as Elon Musk has publicly championed Grok's roast capabilities.

MIT Tech Review

The dominant paradigm of measuring AI progress by pitting models against humans on isolated tasks is increasingly seen as misleading and inadequate, according to researchers pushing for a rethink. Critics argue that benchmark saturation, data contamination, and the narrow framing of "human vs. machine" comparisons obscure real-world capability gaps and fail to capture how AI systems actually perform in complex, collaborative settings.

TechCrunch AI

Salesforce has announced a sweeping AI-driven redesign of Slack, rolling out 30 new features aimed at making the workplace communication platform a more autonomous, agent-capable environment. The update signals Salesforce's intent to position Slack as a primary interface for AI agents operating across enterprise workflows, not just a messaging tool.

The Verge

OpenAI's ChatGPT can now be used hands-free through Apple's CarPlay dashboard, following Apple's iOS 26.4 update which added support for voice-based conversational apps in the car. The integration marks a significant expansion of ChatGPT's reach into everyday consumer hardware, and opens the door for other AI assistants to compete for in-car voice interaction.

TechCrunch AI

AI recruiting startup Mercor confirmed it suffered a data breach after an extortion hacking group claimed responsibility for stealing data from its systems via a compromised version of the open-source LiteLLM project. The incident highlights the growing supply-chain security risks facing AI companies that depend on shared open-source infrastructure.

Google AI

Google has made Veo 3.1 Lite available in paid preview through the Gemini API and Google AI Studio, positioning it as the most affordable entry point in its video generation model lineup. The release continues Google's push to make high-quality generative video accessible to a broader range of developers and enterprise customers.

TechCrunch AI

AI video company Runway is launching a $10 million fund alongside a Builders program designed to support early-stage startups that use its models as a foundation. The move reflects Runway's ambition to evolve beyond a creative tool into a platform company, with a focus on real-time "video intelligence" applications.

The Verge

Generative AI is creating deep ideological and practical rifts inside art and design schools, as faculty, students, and administrators clash over whether to embrace the tools or resist them in the name of preserving craft. For students training in fields like 3D modeling and animation, the stakes are existential — the jobs they're preparing for are already being disrupted before they graduate.