In this episode, we cover a severe security incident at **Meta** where an internal AI agent exposed sensitive data. We also look at **OpenAI's** strategy to consolidate its tools into a new desktop "superapp" and **Perplexity's** entry into the medical space with Perplexity Health.

Also in the news: **WordPress** gives AI agents write access to sites, and **Meta** deploys AI for content moderation.

Sources:
• ITPro: https://www.itpro.com/security/meta-engineer-trusted-advice-from-an-ai-agent-ended-up-exposing-user-data
• CNET: https://www.cnet.com/tech/services-and-software/openai-plans-to-combine-its-ai-tools-in-a-desktop-superapp/
• TNW: https://thenextweb.com/news/perplexity-has-launched-perplexity-health
• TNW: https://thenextweb.com/news/wordpress-com-lets-ai-agents-write-publish-and-manage-your-site
• The Register: https://www.theregister.com/2026/03/20/meta_ai_moderation/

In this episode of **AI RADAR**, the landscape of defense AI shifts as the **Pentagon drops Anthropic** in favor of OpenAI, sparking immediate user backlash over the military contract. Meanwhile, a federal court hands **Perplexity AI** a major victory against Amazon, preserving the future of "Agentic Commerce." Plus, **Nvidia** restarts operations in China and integrates Groq tech to dominate the chip market.

Also on the radar:
• OpenAI drops cost-effective **GPT-5.4 Mini and Nano** models.
• Google permanently axes its risky "What People Suggest" health feature.

Sources:
• Cybernews: https://cybernews.com/ai-news/pentagon-dumps-anthropic-fast-tracks-openai-grok/
• SiliconANGLE: https://siliconangle.com/2026/03/18/court-rules-perplexitys-ai-bots-can-stay-amazon/
• Financial Times: https://www.ft.com/content/nvidia-restarts-china-chips-manufacturing
• AASTOCKS: http://www.aastocks.com/en/stocks/news/aafn-con/NOW.1330000
• The Guardian: https://www.theguardian.com/technology/2026/mar/16/google-scraps-ai-search-feature-health

On today's episode of AI RADAR, we break down the massive announcements from **Nvidia GTC 2026**. CEO Jensen Huang predicts over **$1 trillion** in AI infrastructure revenue and unveils the new **Vera Rubin** architecture alongside an OS for agentic AI.

Next, **Google DeepMind** launches a cognitive framework to finally define and benchmark **AGI**, moving beyond "vibes" to scientific measurement. Then, a historic class-action lawsuit hits Elon Musk's **xAI**, with teens alleging **Grok** was used to generate deepfakes.

Plus, **Mistral AI** releases "Forge" for custom models, and **Meta** delays "Avocado" due to performance gaps.

Sources:
• VentureBeat (Nvidia): https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF2n8ayyppyZrLYI4BlufQiAtTlt5JLAtl3ZWRDT5barQRGYoAy2fL3WCSNGwiHSD_KASIkbWxDjBh9w8aaFc-q3B6PYDQXRxcUNw1kXIUqiz4dtvEZtma0NvLLgDehkMyrNE4IeG9wEA4uYXYuI_nrHI1IUWCmQQzJQctzOpPetjfhp9SoRfPjG0bAi0izcMuP9ZJvY5zjZJrpYhyTWOtl
• The Tech Buzz (DeepMind): https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFch2We0JGZ3-pX04i5R_4PNTLY8pcI67uzsTo4SFYtS7v7l5slhWtBCx5FztCDAy0sxjRsGhc9tGJibD-Tx9v4haK89eJ_x5FdIyzQAXJcOnzeQNjSZL0Xy0Mi-iG2NH6442jl7eUCywsRRT13Gpx7H2s-PX5SehF8slSW87C-esMb4Mx1gd9WWaajpOo8ew4dcj60v46BPZOKpVyN
• The 19th News (xAI): https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQF7xF-u4tJzp0zB0hd6L2HIU_tLC5RW7lwl37SdD30Va34Pln81Xo5CsRbe4WoS1qL8vaQDCJGeO5FqGGzQaz0EDSRqw7ND6H1VmLderdeCgUhmppzvBZzKU9uWESPIkmO1nK8o1dln-Qlf3Rb29GGrvk7SKjeDUPH5ld1i280

In this episode of **AI RADAR**, we break down the massive industry shift towards "Agentic AI" as Nvidia unveils **NemoClaw** to bring security to autonomous enterprise agents. We also analyze reports that **OpenAI** is deprioritizing side projects to focus strictly on coding tools and enterprise solutions ahead of a potential IPO.

Meanwhile, **Meta** is reportedly planning a 20% workforce reduction to fund a projected **$135 billion** AI infrastructure expansion in 2026. Plus, Anthropic doubles usage limits for off-peak users, and Google quietly removes a controversial AI health search feature.

Sources:
• ZDNET: https://www.zdnet.com/article/nvidia-devises-a-more-secure-openclaw-stack-for-enterprises-how-nemoclaw-works
• Investing.com: https://www.investing.com/news/stock-market-news/openai-working-on-shifting-focus-to-core-coding-enterprise-business-wsj-3928123
• PCMag: https://www.pcmag.com/news/meta-reportedly-plans-to-lay-off-20-of-its-staff-amid-ai-push
• PCWorld: https://www.pcworld.com/article/2261123/anthropic-is-doubling-claude-ai-limits-during-off-peak-hours.html
• The Guardian: https://www.theguardian.com/technology/2026/mar/16/google-scraps-ai-search-feature-medical-advice

On this episode of AI RADAR: **Meta** is reportedly planning to cut 20% of its global workforce to fund a massive $600B investment in AI infrastructure and its "Avocado" model. Meanwhile, the **US Government** has dropped strict global AI chip export rules, handing a major regulatory win to Nvidia. Plus, **Google DeepMind** unveils "AlphaEvolve," an AI system that autonomously solved a decade-old math problem regarding Ramsey numbers.

Also: **OpenAI** confirms ChatGPT ads are staying in the US for now, and **Nvidia** prepares to showcase "Agentic AI" at GTC.

Sources:
• HRKatha: https://www.hrkatha.com/news/layoff/meta-to-trim-workforce-by-20-amid-ai-push/
• 24/7 Wall St: https://247wallst.com/technology-3/2026/03/15/government-drops-sweeping-ai-chip-export-rules-can-nvidia-start-growing-again/
• 36Kr: https://36kr.com/p/3195246738554245
• Bleeping Computer: https://www.bleepingcomputer.com/news/technology/openai-says-chatgpt-ads-are-not-rolling-out-globally-for-now/
• Investing.com: https://www.investing.com/news/stock-market-news/nvidia-to-focus-on-competitionbeating-ai-advances-at-megaconference-3337965

The AI hardware war intensifies as **Meta** reportedly considers shifting to **Google TPUs** to reduce reliance on Nvidia, shaking up the market. Meanwhile, **Nvidia** strikes back, securing its future infrastructure with a massive **$2 billion investment in Nebius Group** to deploy the next-gen **Rubin platform**.

In the courtroom, the judge overseeing **Elon Musk’s lawsuit against OpenAI** expresses doubt over the **$134 billion** damages claim but allows the high-stakes trial to proceed to April 2026.

Also on the radar: **OpenAI** aims for 1 billion users with new video scaling, and the **Pentagon** orders the immediate removal of **Anthropic** AI from military systems.

Sources:
• Simply Wall St: https://simplywall.st/stocks/us/software/nasdaq-nbis/nebius-group/news/nebius-group-nbis-is-up-264-after-nvidias-2-billion-ai-par
• Times of India / FT: https://timesofindia.indiatimes.com/technology/tech-news/after-saying-we-are-fine-for-months-nvidia-seemingly-accepts-that-google-and-meta-are-coming-for-it/articleshow/119024345.cms
• Times of India: https://timesofindia.indiatimes.com/technology/tech-news/based-on-what-ive-seen-judge-doubts-elon-musks-134-billion-damages-claim-against-openai/articleshow/119022345.cms
• AI CERTs: https://aicerts.io/news/openais-billion-user-push-and-video-ai-scaling-strategy
• CBS News: https://www.cbsnews.com/news/internal-pentagon-memo-orders-military-commanders-remove-anthropic-ai-technology-key-systems/

Big moves today: **Nvidia** invests **$2 billion** in Nebius for AI cloud infrastructure, while **Anthropic** launches the **Claude Marketplace** to rival OpenAI's ecosystem. Plus, **Yann LeCun's** AMI Labs raises **$1.03 billion** to build "World Models" that understand physical laws.

Also: **OpenAI** acquires Promptfoo for security, and **Alibaba Cloud** updates Qwen for SE Asia.

Sources:
• Nvidia's $2B Bet: https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGLCskSogTLriv2gGHVxWQ7rVr-GAkZwotZ6I4wVwz7F6U5_W6LQT2EPqgZZl3Q_FgDqxWE-HNPqy_aOllF0z90JLqiqjJe8Xwy5NkdaZpVu8zAhNLiMK8KFD5YyPC_B_gt9dxLUjnqDHsjIF5o6jgXFMPZhsB23711bvaNU3NyzNYZx2NC8Z6KJ5GSb-EOwjq6Xx2ibPwAG4klPcR_APHxrEilHLDa_4za7UWceX8f_IWS8tw=
• Anthropic & OpenAI News: https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEMjVW-W2tckW4vo6kooeh3cM-jOqvBzbSJmTKdJJwUfSOWETj9IeEUp_GtwWOIDB8wEXm1oDHVz6kG5TaUC93-uW49vUgdtpmt2seU6_WmP76M086RYM6KRUcxwWk8UQ-xZ2EWigPeSRx7TNE94XVcMzMnYXIkGylnyQ95wKk=
• LeCun's AMI Labs: https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHUGBpNq3bdz6FM-Nb8i2S_B18cgZlLb2OyBaXjnG9ZluPS6zhIbOgyruQRSdFPEhC9GzX_HKB6zXqQvfkOBaW9dIjPfTZ7hjO2GnB1CigAwqKFhh8awpbN0ov70GjP9sOw8yXjIryR8xlyZjZyrxjOm7Zi0FOac2nsT4LumyY=
• Alibaba Qwen: https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGCDJxurXFkhyzxnO4_youNfVzvch6LPrRX_M6dVb435JCYxJJpo5QNKqZrXnudmihPRfwOlC6UiCaReplQ4c8Fvb5bSZpPdYCi67Kx_zpMd-DZO3cl_ILn3kpf3uT8XO7MOReDokaZwLtHD_zQDUMrx3UY0HB3idrMoDpHMxTfivj95sTWr8sh-njRC-3efjksRj0eh-WEhg==

On this episode of AI RADAR: **Meta** reportedly delays its "Avocado" model release after benchmarks showed it failing to beat Google's **Gemini 3.0**, raising questions about Zuckerberg's $135B compute strategy. Meanwhile, a WSJ investigation reveals **ByteDance** is bypassing U.S. sanctions to access 36,000 **Nvidia Blackwell GPUs** via a "cloud loophole" in Malaysia.

Also, **Nvidia** challenges its own customers with a strategic $26 billion pivot into **open-weight models**, and **OpenAI** releases new safety protocols to stop prompt injections.

Sources:
• The Business Times: https://www.businesstimes.com.sg/companies-markets/telcos-media-tech/meta-delays-rollout-new-ai-model-after-performance-concerns
• Investing.com: https://www.investing.com/news/stock-market-news/chinas-bytedance-gets-access-to-nvidia-blackwell-chips-wsj-3929481
• Wired: https://www.wired.com/story/nvidia-open-weight-models-investment/
• CTV News: https://www.ctvnews.ca/business/nvidia-to-invest-us-2-billion-in-neocloud-nebius-amid-ai-data-center-push-1.7228834
• OpenAI: https://openai.com/index/improving-instruction-hierarchy/

On today's episode, the AI hardware wars escalate as **Nvidia invests $2 billion** in Nebius Group to build European "AI Factories," expanding its infrastructure beyond US giants. Meanwhile, **Meta unveils four new custom MTIA chips** on a rapid release cycle, signaling a major move to cut reliance on Nvidia for inference workloads.

Also, the **US State Department has officially dropped Anthropic's Claude** following a directive labeling the model a "supply chain risk" due to disagreements over autonomous weapons usage.

Finally, we touch on a global **Claude service outage** and OpenAI's new research on **securing AI agents** against prompt injection.

Sources:
• Seeking Alpha: https://seekingalpha.com/news/4357602-nebius-surges-after-nvidia-plans-2b-investment-under-collaboration
• SiliconANGLE: https://siliconangle.com/2026/03/11/meta-debuts-internally-developed-ai-chips-inference-workloads/
• TechRadar: https://www.techradar.com/computing/artificial-intelligence/anthropics-claude-models-are-no-longer-available-us-state-department-ditches-claude-on-the-orders-of-trump-while-senate-approves-gemini-chatgpt-and-copilot-for-use
• 9to5Mac: https://9to5mac.com/2026/03/11/claude-ai-and-code-are-experiencing-log-in-issues-and-slow-performance/
• OpenAI: https://openai.com/index/designing-ai-agents-to-resist-prompt-injection/

On this episode of AI RADAR, the clash between ethical AI and national security reaches a boiling point.

First, **Anthropic sues the Pentagon** and the Trump administration after being blacklisted for refusing to allow its models to be used for autonomous lethal weaponry. Next, **Meta acquires Moltbook**, a viral social network for AI agents, signaling a major push into multi-agent systems to compete with OpenAI. Then, a federal judge grants **Amazon an injunction against Perplexity**, blocking its "Comet" browser agent from unauthorized shopping transactions.

Plus, **Nvidia backs Mira Murati's** new startup with massive compute power, and **Google Gemini** gets deeper integration into Workspace to auto-draft documents.

Sources:
• AP News: https://apnews.com/article/anthropic-sues-trump-supply-chain-risk-pentagon-2b8b3
• Axios: https://www.axios.com/2026/03/10/meta-acquires-moltbook-ai-social-network
• Search Engine Land: https://searchengineland.com/court-restricts-perplexity-ai-shopping-bot-amazon-438200
• PYMNTS: https://www.pymnts.com/artificial-intelligence-2/2026/nvidia-invests-in-mira-muratis-thinking-machines-lab/
• Google Workspace: https://workspace.google.com/blog/product-announcements/gemini-update-reimagines-content-creation