One-Line Summary

Washington sketches a national AI rulebook, OpenAI exits viral video with Sora, and investors pour billions into vertical AI while Big Tech trims headcount.

Big Tech

OpenAI Shuts Down Sora, Its Viral AI Video App

OpenAI, the company behind ChatGPT, discontinues Sora, its short‑form AI video social app launched last fall, citing a pivot and promising guidance to preserve user creations. The shutdown follows months of controversy over deepfakes and nonconsensual content, and comes just days after a post outlining new safety standards. ¹

Beyond safety, the business math looks rough: AI video is compute‑hungry and costly at scale, and the app lacked professional‑grade editing or a solid ad model. Multiple reports say the move blindsides Disney, which had discussed a three‑year, US$1B character licensing tie‑up that ultimately never closed; insiders add that Sora drained compute from other teams. The decision signals a refocus on enterprise products, coding, agents, and a consolidated “super app.” ²

Analysts frame Sora’s end as a strategic retreat: OpenAI avoids the messy economics and liability of running a social network while chasing higher‑margin enterprise workflows. Commentary notes Sora’s virality but also its “AI slop” problem and moderation burden—an arena even social incumbents struggle to police. If you built workflows around Sora, expect migration tools, but plan alternatives now. ^{3 4 5}

Industry & Biz

White House Releases National AI Policy Framework

The White House publishes a National Policy Framework for AI—a legislative blueprint urging Congress to create a minimally burdensome national standard with targeted preemption of certain state AI laws. It prioritizes seven areas: child safety, community safeguards (including ratepayer protection from data center costs), IP and digital replicas, free speech protections, innovation (regulatory sandboxes, federal datasets), workforce skills, and federal preemption boundaries. No new “AI super‑regulator” is proposed; sector regulators would lead. ⁶

Critically, the framework favors preempting state AI laws that create “undue burdens,” while preserving state police powers for general consumer and child protection, zoning for AI infrastructure, and states’ own AI procurement rules. On copyright, the administration signals that training on copyrighted works need not violate law, but recommends letting courts hash out fair‑use boundaries and exploring non‑mandatory collective licensing. For “digital replicas” (voice/likeness), it backs a federal right with clear First Amendment carve‑outs. ⁷

Politically, the blueprint contrasts with Sen. Marsha Blackburn’s 291‑page TRUMP AMERICA AI Act, which is more prescriptive on duties of care, audits, and liability. The framework’s lighter touch may be easier to negotiate across committees (e.g., Senate Commerce) but leaves thorny topics—like Section 230 or mandatory audits—for later. Companies should note: existing state AI laws (e.g., CA SB 53, CO AI Act) remain enforceable until courts or Congress say otherwise. Plan for dual compliance in the near term. ^{8 9}

Legal AI Startup Harvey Hits $11B Valuation on $200M Raise

Vertical AI is having a moment: Harvey, which builds AI tools for legal and professional services, raises $200M at an $11B valuation. Its products span contract analysis, compliance, due diligence, and litigation support, reportedly used by 100,000+ lawyers across 1,300 organizations, with ARR climbing to $190M in January from $100M last August. Backers include GIC and Sequoia. ¹⁰

Harvey’s rise shows defensible value in “applied AI” where accuracy, workflow fit, and domain data matter as much as raw model performance. Think of it like Salesforce during the cloud shift: the moat is go‑to‑market, integrations, and trust, not just model weights. Clients like NBCUniversal and HSBC suggest cross‑industry pull beyond law firms. ¹⁰

For buyers, the signal is clear: specialized AI copilots and agents are becoming budget line‑items, not experiments. Expect faster procurement cycles where vendors can quantify time‑to‑value (e.g., hours saved in redlining) and address risk (audit trails, confidentiality). If you’re building in a vertical, this is validation to deepen workflows rather than chase general chat. ¹¹

Oracle Makes Significant Layoffs Amid Massive AI Infra Spend

Oracle begins a significant round of layoffs—reports and employee posts suggest thousands globally—as the company leans into debt‑funded AI infrastructure, including a reported $50B debt raise and participation in the $500B “Stargate” data center initiative with OpenAI and others. Cuts reportedly affect cloud, database, and NetSuite groups, with some roles cited as shrinking due to AI. Oracle had ~162,000 employees as of May 2025. ¹²

The strategy: trim opex while scaling capex to meet surging AI compute demand from customers like Nvidia, Meta, OpenAI, TikTok, and xAI. Investors worry about the “SaaSpocalypse” narrative—AI cannibalizing traditional software—but Oracle argues it uniquely benefits as an AI infra supplier. Shares are volatile; the firm touts cost discipline to reassure markets. ¹³

For talent and vendors, read this as a reshuffle—not a retreat. Infra buyers are consolidating around hyperscale‑like footprints, and partners with GPU capacity, grid access, and capital markets agility will set the pace. Expect continued hiring in data center ops, power, and networking, even as software roles get rationalized. ^{14 15}

Kleiner Perkins Raises $3.5B to Double Down on AI $

Kleiner Perkins closes 3.5B across two funds—$1B for early‑stage (KP22) and $2.5B for growth—marking a step‑up from ~$2B raised in 2024. The firm cites the “AI super‑cycle,” with recent leads in Together AI, Harvey, OpenEvidence and exposure to Anthropic. It’s also benefited from exits like Figma and Brex. ¹⁶

The takeaway: dry powder for AI remains abundant, especially for startups pairing strong model access with deep workflow integration and early revenue traction. Expect bigger late‑stage checks to scale go‑to‑market and compliance, not just compute. ¹⁷

For founders, this favors clear ICPs, enterprise‑ready security/compliance, and credible path to gross margin expansion (e.g., retrieval, fine‑tuning, usage controls). For buyers, anticipate more polished options and competitive pricing as VC‑backed vendors fight for category leadership. ¹⁸

Community Pulse

Hacker News (18↑) — Concern that federal preemption and liability limits could shield AI developers from accountability, weakening state enforcement tools. ⁹

Hacker News (50↑) — Users debate why viral AI apps get shut down instead of spun off or sold.

"Serious question, but when companies like OpenAI and Google roll out something like this they likely get millions of users overnight. I get why it's a distraction, money sink and they don't want to work on it. But presumably it's worth something? The user base alone. Why don't they ever sell these things? Keep a good chunk of the equity w/ some exclusive deals to use their models, but spin it off. How much could a social network like this be worth?" ²

What This Means for You

If you operate in multiple U.S. states, the White House framework—if enacted—could simplify AI compliance by preempting conflicting state rules, while leaving consumer and child‑protection enforcement intact. Practically, that means you’ll still need robust disclosures, safety features, and bias mitigation for consumer‑facing AI in the near term. Dual‑track your compliance: current state laws plus a future federal baseline. ^{7 9}

For marketing and media teams, Sora’s shutdown underscores the fragility of building on viral, consumer‑facing AI platforms. Reassess content strategies dependent on third‑party AI video feeds and prioritize tools with enterprise SLAs, editing controls, and provenance features. Budget for compute‑linked costs and content‑moderation overhead if you run user‑generated AI features. ^{3 4}

If you’re buying or building AI for specialized workflows (legal, finance, healthcare), Harvey’s funding wave signals a durable shift toward vertical copilots and agents. Expect faster product velocity, richer integrations, and more vendor diligence on security and audit trails. Use pilots with measurable KPIs (time‑to‑first‑draft, review accuracy) to justify 2026 renewals. ¹⁰

For tech leaders, Oracle’s layoffs and capex plans foreshadow capacity constraints and consolidation in AI infra. Lock in GPU/compute commitments and explore hybrid strategies (on‑prem + cloud) where latency, data residency, or cost predictability matter. Expect vendors to pass through grid and capex costs—negotiate longer terms for price stability. ^{12 14}

Action Items

1. Map dual compliance for AI: List current state AI/consumer rules affecting your product and draft how a federal baseline would change your controls, disclosures, and contracts.
2. Replace Sora‑dependent workflows: Export assets, then pilot two enterprise‑grade AI video tools with editing/provenance to ensure continuity for campaigns.
3. Run a vertical AI pilot: Choose one legal/compliance task (e.g., NDA review) and trial a specialized AI tool with a clear KPI (turnaround time, accuracy) over two weeks.
4. Secure compute capacity: Ask cloud providers about 2H26 GPU availability and pricing; set thresholds to auto‑shift workloads between providers or on‑prem.