#1Meta Fires 700 to Fund $135B AI Machine
Meta cut roughly 700 roles across Reality Labs, recruiting, sales, and Facebook — its second round of layoffs in 2026, following a January cull that hit 1,000+. The explicit logic: shrink headcount to offset the cost of a $135B capex plan for AI infrastructure this year, while simultaneously handing top AI executives stock packages worth up to $921M each.
*Source: [Bloomberg](https://www.bloomberg.com/news/articles/2026-03-25/meta-is-cutting-several-hundred-jobs-amid-record-ai-spending) / [CNBC](https://www.cnbc.com/2026/03/25/meta-layoffs-reality-labs-facebook.html)*
---
#2Google Drops Lyria 3 Pro — AI Music Goes to Full Songs
Google DeepMind released Lyria 3 Pro, upgrading from 30-second clips to full 3-minute structured tracks with controllable intro/verse/chorus/bridge. It's rolling into the Gemini app (paid tier), Google Vids, and a newly acquired tool called ProducerAI, plus Vertex AI and the Gemini API. All output is watermarked with SynthID.
*Source: [TechCrunch](https://techcrunch.com/2026/03/25/google-launches-lyria-3-pro-music-generation-model/) / [Music Business Worldwide](https://www.musicbusinessworldwide.com/watch-out-epidemic-sound-google-launches-lyria-3-pro-ai-model-that-can-generate-3-minute-tracks/)*
---
#3OpenAI Opens the Bounty on Its Own Models
OpenAI launched two new bug bounty programs: a general Safety Bug Bounty (via Bugcrowd) targeting prompt injection, agentic hijacking, and MCP exploits, and a separate GPT-5 Bio Bug Bounty offering $25K to the first researcher who can crack a 10-level biosecurity jailbreak challenge. This signals OpenAI is treating safety risks as a security surface worth paying for.
*Source: [OpenAI](https://openai.com/index/safety-bug-bounty/) / [Cybersecurity News](https://cybersecuritynews.com/openai-safety-bug-bounty/)*
---
#4Google Personal Intelligence Expands to Free Users
Google is rolling out its Personal Intelligence feature — which lets Gemini read your Gmail, Photos, and YouTube history — to all US users, not just paid subscribers. A quiet but significant expansion of how much ambient data Google's AI can access and act on.
*Source: [Crescendo AI News](https://www.crescendo.ai/news/latest-ai-news-and-updates)*
---
#5OpenAI Shelves Erotic Chatbot Indefinitely
OpenAI has halted development of an adult content chatbot product, the Financial Times reports. No official reason given, but the move comes amid heightened scrutiny of AI safety and the company's Safety Bug Bounty launch — suggesting internal prioritization shifted.
*Source: [Radical Data Science / FT via](https://radicaldatascience.wordpress.com/2026/03/26/ai-news-briefs-bulletin-board-for-march-2026/)*
---
#6Qwen 3.5 and DeepSeek V4 Hit the Frontier
Alibaba's Qwen 3.5 and DeepSeek V4 both dropped in March as multimodal, agentic-capable models pushing into territory previously dominated by OpenAI and Google. Open-source options like Mistral Small 3.1 24B and Kimi K2.5 also shipped this week, keeping pressure on closed providers.
*Source: [DEV Community](https://dev.to/aibughunter/the-llm-and-ai-agent-releases-that-actually-matter-this-week-march-2026-5d7i) / [ClawPod](https://www.clawpod.co/blog/llm-news-open-source-ai-models-march-2026)*
---
#7Meta and Google Face Legal Reckoning Over Child Harm
Two separate jury verdicts found Meta and Google liable for harms to minors — one Los Angeles case awarded $6M after a plaintiff argued Instagram and YouTube contributed to depression and suicidal ideation. These aren't the first, but the scale and specificity are escalating fast, and could reshape platform liability law.
*Source: [Reuters via Motley Fool](https://www.fool.com/coverage/stock-market-today/2026/03/26/stock-market-today-march-26-nasdaq-falls-2-4-after-meta-and-micron-drop-sharply/)*
---
**Big Picture**
Today's news is a portrait of an industry doubling down while the walls close in from multiple directions. The same companies laying off workers are projecting record AI spend; the same models being released to the public are now being offered up for red-teaming bounties. The safety-and-control discourse has moved from policy papers to cash prizes and courtrooms. The race to deploy is accelerating, but so is the infrastructure of accountability — legal, technical, and regulatory. The question isn't whether AI is getting more capable; it's whether the guardrails are keeping pace.
