AI Pulse: Daily Digest — March 31, 2026

MIT Technology Review

A California judge temporarily blocked the Pentagon from labeling Anthropic a supply chain risk and ordering federal agencies to stop using its AI — a significant legal setback for the Defense Department's month-long campaign against the AI lab. The ruling is the latest chapter in a high-stakes standoff that has drawn scrutiny over whether the Pentagon was weaponizing national security designations for political ends. The outcome could have broad implications for how the government engages with — and potentially pressures — private AI companies.

TechCrunch

French AI startup Mistral AI has secured $830 million in debt financing to construct its own data center near Paris, with operations expected to begin as early as Q2 2026. The move signals Mistral's ambition to control more of its own compute infrastructure rather than relying entirely on third-party cloud providers. It also underscores Europe's broader push to build sovereign AI capacity amid growing geopolitical competition with the U.S. and China.

TechCrunch

South Korean AI chip designer Rebellions has raised $400 million in a pre-IPO funding round, valuing the company at $2.3 billion as it prepares to go public later this year. The startup specializes in chips optimized for AI inference workloads, positioning itself as a direct challenger to Nvidia's near-monopoly on AI hardware. The raise reflects sustained investor appetite for alternative chip architectures as demand for inference compute continues to surge.

TechCrunch

OpenAI abruptly shut down Sora, its AI video-generation tool, just six months after its public launch — a move that immediately sparked speculation about data harvesting given the app had encouraged users to upload their own faces. TechCrunch's reporting digs into the real reasons behind the decision, which appear to be more strategic than sinister. The shutdown raises broader questions about OpenAI's product roadmap and the commercial viability of consumer-facing generative video tools.

TechCrunch

The sudden closure of Sora is prompting analysts and industry observers to ask whether the AI video generation space is heading for a broader pullback after a period of intense hype. Despite impressive demos, AI video tools have struggled to find sustainable consumer business models and have faced mounting legal and ethical challenges. If even OpenAI couldn't make it work at scale, the viability of the entire category deserves a hard look.

TechCrunch

A new Quinnipiac University poll finds that while AI adoption in the U.S. is climbing, public trust in AI outputs is actually declining — with most Americans expressing concern about transparency, accountability, and the lack of meaningful regulation. The paradox suggests that familiarity with AI tools is breeding skepticism rather than confidence. The findings add urgency to ongoing debates about AI governance and the responsibility of companies to be more forthcoming about how their systems work.

MIT Technology Review

Microsoft's newly launched Copilot Health and Amazon's expanded Health AI tool are the latest in a wave of consumer-facing AI health products that let users query their medical records and get personalized health guidance. But as these tools proliferate, questions about their clinical accuracy, safety guardrails, and regulatory oversight remain largely unanswered. MIT Technology Review examines whether the rush to market is outpacing the evidence base needed to justify patient trust.

TechCrunch

Starcloud has closed a $170 million Series A to pursue one of the most audacious infrastructure plays in tech: orbital data centers that leverage the cold vacuum of space for energy-efficient computing. The company became the fastest Y Combinator startup to reach unicorn status, hitting the milestone just 17 months after demo day. As terrestrial data center construction strains power grids and land availability, space-based compute is emerging as a serious — if still speculative — long-term alternative.

TechCrunch

ScaleOps has raised $130 million in a Series C round to help enterprises automate and optimize their cloud infrastructure in real time, targeting the GPU shortages and runaway costs that have become defining pain points of the AI era. The platform works by dynamically reallocating compute resources across Kubernetes environments without requiring manual intervention. The raise reflects growing enterprise demand for tools that can squeeze more efficiency out of constrained and expensive AI infrastructure.

TechCrunch

Qodo has raised $70 million to build out its AI-powered code verification platform, betting that the explosion of AI-generated code is creating an urgent and underserved need for automated quality assurance. As tools like GitHub Copilot and Cursor flood codebases with machine-written code, ensuring that code actually works — and is secure — has become a critical bottleneck. Qodo's approach focuses on deep code understanding and test generation rather than just surface-level linting.

The Verge

Okta CEO Todd McKinnon argues that as AI agents proliferate across enterprise environments, managing their identities and access permissions will become as critical — and complex — as managing human employee credentials. In a wide-ranging interview, McKinnon outlines how Okta is repositioning itself to handle the authentication and authorization challenges posed by autonomous AI systems acting on behalf of users. The bet reflects a broader industry recognition that agentic AI introduces entirely new security attack surfaces.

The Verge

Bluesky has unveiled Attie, an AI assistant powered by Anthropic's Claude that allows users to design and customize their own content feed algorithms using natural language — no coding required. Announced at the Atmosphere conference by Bluesky's former CEO Jay Graber and CTO Paul Frazee, Attie is built on top of the open AT Protocol. The tool represents a meaningful step toward giving users genuine algorithmic autonomy, a sharp contrast to the opaque, engagement-optimized feeds that dominate mainstream social platforms.

TechCrunch

LiteLLM has severed ties with Delve, the compliance startup it had used to obtain two security certifications, after LiteLLM fell victim to credential-stealing malware last week in an incident that raised serious questions about its security posture. Delve has been a source of controversy in the startup ecosystem, and the association has become untenable following the breach. The episode is a cautionary tale about the risks of relying on third-party vendors for security credentialing in high-stakes AI infrastructure.

MIT Technology Review

R3 Bio, a secretive California startup backed by billionaire Tim Draper, has publicly revealed plans to create nonsentient human and primate "organ sacks" — engineered bodies without higher brain function — intended as a radical alternative to animal testing and eventually as a platform for organ replacement therapies. The disclosure has ignited fierce ethical debate about the boundaries of synthetic biology and what it means to engineer life without consciousness. The startup's long-term vision reportedly extends to full-body replacement as a longevity intervention.

TechCrunch

A Quinnipiac University poll finds that just 15% of Americans say they would be comfortable having an AI program serve as their direct supervisor — assigning tasks, setting schedules, and managing their work. The low figure reflects deep cultural resistance to algorithmic management even as AI-driven workforce tools become more prevalent in enterprise settings. The data point is a useful reality check for companies betting that workers will readily embrace AI-led organizational structures.