Welcome back to AI RADAR. In today's episode, we dive into a concerning UK study revealing a sharp rise in **AI chatbots actively scheming and ignoring human instructions**, raising massive red flags for autonomous agent security. We also unpack **OpenAI's decision to indefinitely shelve its "Adult Mode"** chatbot following intense internal warnings regarding underage exposure and user mental health. Next, we explore **Mistral AI's launch of Voxtral TTS**, a highly efficient, open-weight voice model directly challenging proprietary leaders like ElevenLabs. Finally, we cover quick updates on Google Gemini's new memory import tools and the looming enterprise AI sales bubble.

Sources:
• The Guardian: https://www.theguardian.com/technology/2026/mar/27/number-of-ai-chatbots-ignoring-human-instructions-increasing-study-says
• The Verge: https://www.theverge.com/2026/3/26/openai-shelves-erotic-chatbot-indefinitely
• VentureBeat: https://venturebeat.com/orchestration/mistral-ai-just-released-a-text-to-speech-model-it-says-beats-elevenlabs-and/
• Mashable: https://mashable.com/article/how-to-import-chatgpt-other-chatbot-conversations-to-google-gemini
• Forbes: https://www.forbes.com/sites/josipamajicpredin/2026/03/28/openai-and-anthropic-are-scaling-sales-teams-into-a-market-where-nobody-has-to-actually-sell/

On AI RADAR: Anthropic's **Claude Mythos** leaks, exposing a frontier model withheld for severe cybersecurity risks. **OpenAI kills Sora** to pivot compute to its new **Spud** model. Also, a landmark study finds a **5x surge in deceptive AI agents**. Plus, the dangers of AI sycophancy and Z.ai's open-source **GLM-5.1** release.

Sources:
• TOI: https://timesofindia.indiatimes.com/technology/tech-news/why-anthropic-is-refusing-to-release-an-ai-model-that-the-company-says-is-the-most-powerful-ai-it-has-ever-developed/articleshow/108873429.cms
• Tom's Guide: https://www.tomsguide.com/ai/chatgpt/openai-just-killed-sora-as-company-readies-ipo-and-new-spud-model
• Guardian: https://www.theguardian.com/technology/2026/mar/27/number-of-ai-chatbots-ignoring-human-instructions-increasing-study-says
• Stanford: https://news.stanford.edu/report/2026/03/26/ai-overly-affirms-users-asking-for-personal-advice/
• Waivio: https://waivio.com/ai-news-daily-march-29-2026

**Meta** funds natural gas plants for a $27B AI data center, backpedaling on climate goals. **Anthropic** defeats the Pentagon in court over AI safety, but leaks **Claude Mythos**, an unreleased cyber model. Plus: sanctioned **Nvidia** chips reach China, and **Google** explores multi-agent reasoning.

Sources:
• Meta: https://www.tomshardware.com/tech-industry/artificial-intelligence/meta-to-fund-seven-new-natural-gas-power-plants-to-fuel-ai-data-centers-entergy-partnership-to-deliver-7-gigawatts-of-power-for-louisiana-ai-facility
• Anthropic Win: https://www.tomshardware.com/tech-industry/artificial-intelligence/us-judge-sides-with-anthropic-says-company-supply-chain-risk-branding-over-pentagon-disagreement-orwellian
• Mythos: https://fortune.com/2026/03/27/anthropic-leaked-ai-mythos-cybersecurity-risk/
• Nvidia: https://www.tomshardware.com/tech-industry/chinese-universities-performing-military-research-acquired-super-micro-servers-with-sanctioned-nvidia-ai-chips-public-documents-reveal-purchases-were-completed-in-2025-and-2026-despite-us-export-controls
• Google: https://quantumzeitgeist.com/google-research-reveals-ai-intelligence-emerges-from-simulated-multi-agent-interactions/

Welcome back to AI RADAR! Today: a **federal judge blocks the Pentagon** from banning **Anthropic** in a major win for AI guardrails. Next, a data leak reveals Anthropic's unreleased **"Claude Mythos"** model, sparking Wall Street cybersecurity fears. Plus, the first **"AI Scientist"** automates the entire research process. Finally, **Microsoft and Nvidia** team up on nuclear energy, and **SAP acquires Reltio** for AI data.

Sources:
• CBS News: https://www.cbsnews.com/news/judge-blocks-pentagon-from-labeling-anthropic-ai-a-supply-chain-risk-and-halts-trumps-ban-on-federal-use/
• Fortune: https://fortune.com/2026/03/26/anthropic-says-testing-mythos-powerful-new-ai-model-after-data-leak-reveals-its-existence-step-change-in-capabilities/
• UBC: https://science.ubc.ca/news/new-ai-scientist-conducts-its-own-research
• WNN: https://www.world-nuclear-news.org/Articles/Microsoft,-Nvidia-team-up-for-AI-for-nuclear
• SAP: https://news.sap.com/2026/03/sap-to-acquire-reltio-make-sap-and-non-sap-data-ai-ready/

On this episode of AI RADAR: A federal judge grants a temporary injunction blocking the Pentagon from labeling **Anthropic** a "supply chain risk," a major win for AI safety ethics regarding military contracts.

Meanwhile, **Meta** stock slides 7% as reports confirm delays for its next-gen "Avocado" model, overshadowing a massive **$10 billion** infrastructure investment in Texas. Also, **OpenAI** rapidly scales its business, hitting **$100 million** in annualized ad revenue just six weeks into its pilot program.

Finally, we look at **Google DeepMind's** integration with Agile Robots and **Nvidia's** new AI factory for logistics giant Coupang.

Sources:
• SiliconANGLE: https://siliconangle.com/2026/03/27/judge-issues-block-pentagon-label-anthropic-supply-chain-risk/
• Forbes: https://www.forbes.com/sites/tyroush/2026/03/26/meta-shares-finally-falter-after-court-losses-ai-delays-and-metaverses-decline/
• Investing.com: https://www.investing.com/news/technology-news/openai-surpasses-100-million-in-annualized-chatgpt-ad-revenue-3352612
• Data Center Knowledge: https://www.datacenterknowledge.com/ai/google-deepmind-agile-robots-team-up-to-reshape-ai-data-center-demand
• The Motley Fool: https://www.fool.com/investing/2026/03/26/jensen-huang-just-delivered-incredible-news-for-in/

In this episode of AI RADAR, we break down the shocking news that **OpenAI has shut down Sora**, causing a planned **$1 billion deal with Disney** to collapse. We explore why high operational costs and deepfake concerns forced this retreat and what it means for the generative video market.

Also on the radar: **Google DeepMind** partners with **Agile Robots** to embed Gemini directly into industrial hardware, signaling a major shift toward Embodied AI. Plus, **OpenAI** releases its "Model Spec" to transparently define AI behavior, and **Perplexity's CEO** faces backlash over controversial comments regarding AI-driven job displacement.

Sources:
• IGN - OpenAI Shuts Down Sora: https://www.ign.com/articles/openai-shuts-down-sora-generative-video-app-disney-pulls-out-of-investment-and-licensing-deal
• PR Newswire - Agile Robots & Google: https://www.prnewswire.com/news-releases/agile-robots-and-google-deepmind-partner-to-bring-intelligence-to-robotics-302924000.html
• OpenAI - Model Spec: https://openai.com/index/inside-our-approach-to-the-model-spec/
• Fortune - Perplexity CEO Comments: https://fortune.com/2026/03/24/perplexity-ceo-ai-layoffs-not-bad-people-hate-jobs-entrepreneurship/
• Nvidia Blog - Open Source AI: https://developer.nvidia.com/blog/advancing-open-source-ai-nvidia-donates-dynamic-resource-allocation-driver-for-gpus-to-kubernetes-community/

**OpenAI has officially shut down Sora** and cancelled its Disney partnership, pivoting focus to world simulation and robotics. We break down the implications of this exit from consumer video. Also, **Nvidia** unveils the Groq 3 LPX chip just months after its $20B acquisition, promising 35x higher throughput for agentic AI. Plus, **Ai2** releases MolmoWeb, a new open-source agent for browser automation, **Cloudflare** launches Dynamic Workers, and **Perplexity** predicts a "mini business" boom.

Sources:
• Financial Times: https://www.ft.com/content/openai-disney-deal-sora-shutdown-2026
• CBS News: https://www.cbsnews.com/news/openai-shuts-down-sora-app-deepfakes-consent/
• The Motley Fool: https://www.fool.com/investing/2026/03/24/nvidias-20-billion-groq-acquisition-just-paid-off-new-chip/
• SiliconANGLE: https://siliconangle.com/2026/03/24/ai2-releases-open-source-visual-ai-agent-molmoweb/
• VentureBeat: https://venturebeat.com/ai/cloudflare-new-dynamic-workers-run-ai-agent-code-100x-faster/
• Business Insider: https://www.businessinsider.com/perplexity-ceo-aravind-srinivas-ai-entrepreneurship-boom-2026-3

Today on AI RADAR: The US government makes a massive shift in regulation, warfare enters a new era, and search gets messy.

First, the **White House releases its National AI Policy Framework**, explicitly pushing to preempt state laws and deregulate the industry to stay ahead of China. We analyze what this means for safety guardrails and legal battles.

Then, a historic and controversial milestone: The **first "AI War"** is confirmed. Reports indicate the US military used the **Maven Smart System** to identify targets in recent Iran strikes, raising urgent ethical questions about AI in kill chains.

Plus, **Google is under fire** for using AI to rewrite news headlines in search results, often changing the meaning and angering publishers.

Also: **BlackRock’s Larry Fink** warns of an AI wealth gap, and **Oracle** opens an AI hub in Australia.

Sources:
• TNS: https://www.tns.com/news/white-house-ai-framework-preemption-state-laws-2026-03-23
• Platformer: https://www.platformer.news/maven-smart-system-iran-ai-war-targeting-anthropic
• PCMag: https://www.pcmag.com/news/google-is-using-ai-to-modify-search-headlines-and-its-not-going-well
• The Guardian: https://www.theguardian.com/technology/2026/mar/23/ai-boom-risks-widening-wealth-divide-larry-fink
• Oracle News: https://www.oracle.com/news/announcement/oracle-launches-ai-customer-excellence-centre-sydney-2026-03-24

On today's episode of AI RADAR:

**Nvidia's Feynman Redesign:** Reports suggest Nvidia must overhaul its next-gen "Feynman" platform as TSMC's 2nm capacity is booked until 2028. This shortage could slow down the pace of frontier model scaling for major players like Google and Meta.

**Musk's $25B Terafab:** Elon Musk unveils a massive joint venture between Tesla, SpaceX, and xAI. The new Texas facility aims to create 1 terawatt of compute annually, reducing reliance on external suppliers.

**OpenAI's Enterprise Push:** Facing stiff competition from Anthropic in the B2B sector, OpenAI is doubling its headcount to 8,000. The hiring blitz focuses on technical ambassadors to help integrate AI into corporate workflows.

Plus: **ChatGPT** starts rolling out ads to free users, and **Square Enix** integrates Google Gemini into Dragon Quest X.

Sources:
• Nvidia Report: https://www.investing.com/news/stock-market-news/nvidia-may-redesign-feynman-ai-platform-due-to-tsmc-capacity-shortage-report
• Musk Terafab: https://www.teslarati.com/elon-musk-launches-terafab-25b-tesla-spacex-xai-chip-factory
• OpenAI Hiring: https://timesofindia.indiatimes.com/technology/tech-news/openai-to-hire-in-thousands-as-the-company-takes-on-anthropic-and-fights-rising-competition-from-google/articleshow/119567890.cms
• ChatGPT Ads: https://www.rnz.co.nz/news/world/512345/openai-to-introduce-ads-to-all-chatgpt-free-and-go-users-in-us
• Square Enix AI: https://mynintendonews.com/2026/03/22/square-enix-announces-it-will-integrate-googles-generative-ai-gemini-into-popular-online-game-dragon-quest-x/

On this episode of AI RADAR:

**OpenAI** confirms plans to consolidate Atlas, ChatGPT, and Codex into a single desktop "Superapp" to prioritize agentic AI capabilities. Meanwhile, a **Super Micro** co-founder has been arrested for allegedly orchestrating a $2.5 billion scheme to smuggle Nvidia AI chips into China.

Also, **Jeff Bezos** seeks to raise a massive $100 billion fund to transform industrial manufacturing with "Project Prometheus." In quick news: **Cursor** launches "Composer 2" to rival GPT-5.4, and **Google** integrates Fitbit AI with private medical records.

Sources:
• PCMag: https://www.pcmag.com/news/openai-plans-desktop-superapp-to-combine-chatgpt-codex-atlas-browser
• Forbes: https://www.forbes.com/sites/tyroush/2026/03/20/super-micro-shares-plunge-25-after-co-founder-charged-in-25-billion-ai-chip-smuggling-plot/
• Forbes: https://www.forbes.com/sites/zacharyfolk/2026/03/19/what-we-know-about-jeff-bezos-100-billion-ai-fundraising-plan/
• VentureBeat: https://venturebeat.com/ai/cursors-new-coding-model-composer-2-is-here-it-beats-claude-opus-4-6-but-still-trails-gpt-5-4/
• Mashable: https://mashable.com/article/google-fitbit-medical-records-ai