AI News Roundup: OpenAI Education Tools, Gemini Workspace, and the Rise of AI Agents

AI news this week reminds us that innovation is moving from pilots in labs to everyday use in offices and classrooms. In Europe, Neura announced what it calls the largest physical AI training center, a bid to accelerate humanoid robotics and keep pace in the global physical AI race. Across the board, executives, researchers, and policymakers are watching how quickly AI shifts from impressive demos to tools that shape decision making, productivity, and skills training.

On the educational front, OpenAI expanded ChatGPT with interactive learning tools that let users manipulate math and science concepts in real time. The feature covers more than 70 core topics, from algebra to physics, and now works across all active user plans. OpenAI reports hundreds of millions of weekly interactions with these tools, underscoring AI’s growing role in formal and informal education. But the week also carried reminders that with opportunity come challenges, as lawsuits and regulatory scrutiny surface alongside a powerful enhancement to how people learn with AI.

Security and governance of AI powered workflows moved to the center of enterprise planning. A prominent security perspective argues that identity should become the control plane for AI agents, with context aware access, zero knowledge credential handling for autonomous agents, and enhanced auditability to track what agents do and under whose authority. This rethink aims to prevent unintended agent actions while preserving the productivity benefits that AI agents can deliver across software development, data analysis, and business processes.

In the productivity arena, Google pushed Gemini deeper into Workspace, enabling the AI to pull data from Drive, Gmail, and Chat to draft Docs, populate Sheets, and assemble Slides. The updates promise faster content creation and smarter cross-file synthesis, with an enterprise Alpha program that lets administrators turn on early access. The broader goal is to turn the Workspace suite into an active knowledge base where AI agents orchestrate complex tasks across multiple apps, reducing the manual digging through documents and emails that used to slow teams down.

Another thread running through the week is the maturation of AI risk management tools. Anthropic and OpenAI have introduced reasoning based vulnerability scanners to complement traditional static analysis tools. These new tools, Claude Code Security and Codex Security, reveal vulnerability classes that pattern matching SAST tools often miss. The takeaway for security leaders is clear: run multiple scanners on representative codebases, establish governance before patches, and prepare for a future where modular AI tools become standard components in secure software supply chains.

Beyond these developments, the week includes strategic corporate moves and policy signals. Meta announced the acquisition of Moltbook, a platform focused on AI agents, signaling a trend toward AI driven social architectures. In publishing and copyright debates, the UK Society of Authors explored human authored labeling to help readers distinguish AI generated works, while in Canada a family sued OpenAI alleging that the company could have prevented a tragic attack. Capital One highlighted a push to build AI ready data ecosystems, illustrating how data governance underpins scalable AI adoption. The industry is also watching the data center race and the entangled questions of AI progress and hype, as observers debate whether the current optimism constitutes a bubble or a necessary phase of rapid evolution.

Amid this flurry, the pace of experimentation continues to accelerate. Andrej Karpathy recently showcased autoresearch, a lightweight open source framework that automates large portions of the research loop. In overnight runs, autonomous agents conducted hundreds of experiments with little human intervention, uncovering insights that would have taken humans far longer to achieve. This glimpse into the future hints at a world where curiosity itself becomes the bottleneck, underscoring the need for governance, ethical guardrails, and thoughtful deployment strategies as AI tools become ever more capable and embedded in everyday work.

Sources

1. Neura launches Europe’s largest physical AI training center: https://aibusiness.com/robotics/neura-launches-physical-ai-training-center
2. Meta acquires AI agent social network Moltbook: https://www.theguardian.com/technology/2026/mar/10/meta-acquires-moltbook-ai-agent-social-network
3. UK Society of Authors logo to identify human-authored books: https://www.theguardian.com/technology/2026/mar/10/uk-society-authors-logo-identify-books-written-by-humans-not-ai
4. OpenAI upgrades ChatGPT with interactive learning tools: https://venturebeat.com/data/openai-upgrades-chatgpt-with-interactive-learning-tools-as-lawsuits-and
5. Anthropic and OpenAI expose SAST reasoning scanners: https://venturebeat.com/security/anthropic-openai-sast-reasoning-scanners-security-directors-guide
6. French AI startup building world models raises 1.03B: https://aibusiness.com/generative-ai/french-ai-startup-building-world-models-raises-1-03-billion
7. Keyboard jamming feature in Guardian: https://www.theguardian.com/money/2026/mar/10/keyboard-jamming-sneaky-way-make-boss-think-working-from-home
8. Google upgrades Gemini for Workspace: https://venturebeat.com/orchestration/google-upgrades-gemini-for-workspace-allowing-it-to-pull-data-from-multiple
9. Family of Tumbler Ridge shooting victim sues OpenAI: https://www.theguardian.com/world/2026/mar/10/tumbler-ridge-shooting-victim-sues-openai-canada
10. Capital One AI ready data ecosystem: https://aibusiness.com/agentic-ai/building-ai-ready-data-ecosystem-at-capital-one
11. Datacenters as targets in warfare: https://www.theguardian.com/technology/2026/mar/10/datacenters-target-warfare-iran
12. Ive taught thousands how to use AI: https://www.theguardian.com/lifeandstyle/ng-interactive/2026/mar/10/teaching-ai-what-i-learned
13. Limits of bubble thinking on AI progress: https://venturebeat.com/technology/the-limits-of-bubble-thinking-how-ai-breaks-every-historical-analogy
14. Autoresearch by Andrej Karpathy: https://venturebeat.com/technology/andrej-karpathys-new-open-source-autoresearch-lets-you-run-hundreds-of-ai