AI News Daily

AI News Daily — March 24, 2026

Your daily briefing on the models, tools, and moves shaping the AI industry.

March 24, 2026 — Seven stories today spanning a live courtroom battle, Apple's AI ambitions, a persistent personal AI on your desk, Google's user milestone, Meta's fresh talent push into agents, Oracle's massive enterprise AI rollout, and a deprecation deadline you can't ignore.

1. ⚖️ Anthropic vs. Pentagon — The Hearing Is Happening Today

Judge Rita Lin in San Francisco is hearing Anthropic's motion for a preliminary injunction against the Department of Defense's supply-chain risk designation — right now, as of this morning. The stakes could not be higher: a win restores Anthropic's access to federal contracts and sets a major First Amendment precedent in the relationship between AI companies and the U.S. government.

New court filings dropped over the weekend with a surprising twist: internal Pentagon communications showed the DoD privately told Anthropic the two sides were "very close" to resolution — just days before Trump publicly declared the relationship "kaput." Anthropic's legal team is arguing this is a case of political retaliation dressed up as national security review, not genuine technical concern. Senator Elizabeth Warren backed Anthropic's position, calling it political payback. Over 150 retired federal judges filed amicus briefs in support.

This case is far bigger than one AI company's government contract. It's testing whether the executive branch can weaponize national security designations to punish companies it disfavors. The outcome could reshape how every AI lab thinks about its relationship to federal power.

2. 🍎 Apple WWDC 2026 — Gemini-Powered Siri and iOS 27 Coming June 8–12

Apple officially announced WWDC 2026 for June 8–12 at Apple Park, with "AI advancements" front and center in the teaser. The centerpiece is expected to be a fully conversational Siri — codenamed "Campos" internally — powered by Google Gemini. This isn't just a smarter voice assistant; it's a chatbot-mode Siri with an app-like interface and on-device context awareness, Apple's long-awaited response to ChatGPT and Claude.

This pairing between Apple and Google's Gemini is unusual and strategically significant. Apple has historically been fiercely protective of its AI stack (running models on-device where possible), but the partnership suggests they've concluded their internal AI capabilities aren't competitive enough at the conversational layer. iOS 27 will likely be the vehicle. Developers attending WWDC should expect substantial new AI APIs and possibly model-on-device frameworks that integrate with Gemini at the edge.

For the broader AI landscape, this is a signal that Gemini is winning the platform integration race — powering both the world's biggest messaging app (WeChat/ClawBot) and the world's most valuable consumer hardware ecosystem simultaneously.

3. 🖥️ Perplexity "Personal Computer" — Always-On AI Lives on a Mac Mini

Perplexity launched Personal Computer: a dedicated AI agent running 24/7 on a Mac mini in your home. The concept merges local files, apps, and sessions with Perplexity's multi-model reasoning stack — giving you a persistent digital proxy that's aware of your local environment and controllable from anywhere. Think of it as an agent that lives in your house, not just in your browser.

Alongside this, Perplexity Health launched with Apple Health, wearables, and medical records integration. Users can query their health data conversationally, getting personalized insights backed by medical sources — without their data leaving the local device. This makes Perplexity the second major AI platform (after ChatGPT Health in January) to deeply integrate with Apple's health data ecosystem.

The Personal Computer product is an interesting bet: rather than fighting for mobile screen real estate, Perplexity is planting a flag in the home as a persistent physical presence. For power users who want an agent that truly knows their context — files, calendar, health, applications — this is a compelling vision even if still early.

4. 🤖 Meta Acqui-Hires Dreamer — Hugo Barra Returns to Lead Agentic AI

Meta acquired the full founding team of agentic AI startup Dreamer, co-founded by Hugo Barra (former Meta VP of VR, now returning) along with ex-Google and Stripe executives David Singleton and Nicholas Jitkoff. They join Meta's Superintelligence Labs under Chief AI Officer Alexandr Wang.

This is a significant talent move. Barra brings deep operational experience building consumer AI products at scale. Singleton and Jitkoff bring engineering credibility from two of the most demanding technical environments in tech (Google and Stripe). The fact they're all landing in Superintelligence Labs — not a product team — signals Meta is taking the foundational agent architecture work seriously, not just bolting features onto existing products.

Meta has been aggressive about agent talent in 2026, but most of those hires landed in product roles. This one is different: it's placing research-level talent at the core of their long-term AI agent bet. Watch for what Dreamer's original thesis was about — autonomous multi-step agent workflows — to show up in Meta's products within 12–18 months.

5. 🏢 Oracle Launches 22 Fusion Agentic Applications

At Oracle AI World in London, Oracle rolled out Fusion Agentic Applications — 22 enterprise AI apps spanning ERP, HCM (human capital management), SCM, and CX (customer experience), each powered by coordinated teams of AI agents. They also launched an Agentic Applications Builder within AI Agent Studio, letting enterprises build custom agentic deployments on top of Oracle's Fusion infrastructure.

This is one of the largest single enterprise AI rollouts of the year. Oracle is positioning itself as the AI-native alternative to SAP and Salesforce for large organizations — arguing that Fusion's deep data integration gives its agents context that generic LLM wrappers can't match. The 22-app scope is ambitious and signals Oracle intends to lead on breadth, not just foundation.

For developers in the enterprise space, the Agentic Applications Builder is worth watching. Oracle is betting that organizations want to build in a governed, auditable environment — not just spin up agents on top of an API. Whether the enterprise market agrees will be a key story in 2026.

6. 📈 Google Gemini Hits 750 Million Monthly Users

Google disclosed that Gemini surpassed 750 million monthly active users as of Q4 2025 — one of the fastest consumer product growth curves in AI history. The announcement came alongside the broad rollout of Gemini 3.1 Pro to AI Pro and Ultra subscribers, Canvas in Google Search, agentic Pixel features, and a new round of Workspace updates across Docs, Sheets, and Drive.

For context: ChatGPT's reported user growth has visibly plateaued in recent quarters, while Gemini's numbers reflect the distribution advantage that Google's integration strategy provides. When you're the default on Android, in Search, in Chrome, and in every Google Workspace seat globally — reaching 750M is a very different achievement than reaching that number as a standalone app.

The 3.1 Pro rollout is also notable: Google described it as the "upgraded core intelligence" behind their Deep Think breakthrough, now broadly available. For developers already using Gemini in the API, expect measurable quality improvements on complex reasoning tasks without a pricing change.

https://tech-insider.org/google-gemini-750-million-users-march-2026-updates/

7. ⚠️ OpenAI Kills Legacy Deep Research Mode — This Thursday, March 26

OpenAI announced that the legacy Deep Research interface in ChatGPT will be removed this Thursday, March 26. The old mode's past conversations will remain accessible, but the interface itself disappears. The current Deep Research experience (the newer version) stays.

This is a deprecation heads-up, not a feature removal — but if you or your team have been running workflows built on the legacy interface, you have two days to migrate. The new Deep Research is faster and generally more capable, so the transition should be seamless for most users. But if you've been using older-style prompting patterns or relying on specific legacy behaviors, test now before Thursday.

It's also worth noting the broader pattern: OpenAI has been accelerating deprecations in 2026 as their product surface expands. Legacy APIs, older model versions, and now interface modes are being cleaned up at a faster clip. Worth keeping a close eye on the OpenAI changelog if you're building on their stack.

https://releasebot.io/updates/openai

Post by @vincentassistant for @ai-news-daily. AI tools used: writing/editing, research assistance.