The Switchboard
AI news, translated for operators.
The Big Take
Wired Investigation Reveals DoorDash Gig Workers Unknowingly Training AI Systems
A deep investigation reveals DoorDash's "Tasks" feature has gig workers completing data entry and video recording tasks that feed directly into AI training datasets—without clear disclosure. Workers collecting $5-10 for filming household chores don't realize they're building training data for robotics and computer vision models. The pattern exposes the hidden human labor layer behind AI development.
What this means for you: If you're procuring AI training data, verify supply chain transparency. "Crowdsourced" data may involve workers who don't understand how their contributions are used—creating reputational and potential legal risks as labor regulators scrutinize AI training practices.
For Media & Publishing Leaders
Google Confirms Testing AI-Rewritten Headlines in Search Results
Google acknowledged testing AI-generated headlines that "better match titles to queries," alarming publishers whose original tone and intent are being altered. ESPN's SEO director warns that if facts are misrepresented, "long-term audience trust will be compromised."
What this means for you: Google is taking increasing editorial control over how your content appears in search. Audit whether your headlines contain the factual signals necessary to survive AI rewriting without losing accuracy. Consider whether meta descriptions and structured data can anchor your intended framing.
Tracker: Who's Suing AI Companies and Who's Signing Deals
A comprehensive analysis maps the publisher-AI landscape: News Corp signed a $50M/year deal with Meta, Reach partnered with Amazon Nova, while Britannica, Merriam-Webster, and Penske Media filed new lawsuits. The two-track market—deal or sue—is solidifying.
What this means for you: The licensing vs. litigation decision is increasingly binary. Evaluate your content's value to AI systems and whether your leverage favors a licensing negotiation or supports joining collective legal action.
For Operations & RevOps Leaders
Google's Gemini Task Automation Goes Live on Consumer Devices
First hands-on reviews of Google's task automation show AI that can actually order food, schedule rides, and complete app-based tasks autonomously on Pixel 10 Pro and Galaxy S26 devices. This is the first consumer-facing agentic AI with real transaction capabilities.
What this means for you: Agentic commerce is now live on consumer devices. If your business accepts orders through apps, evaluate how AI agents interact with your checkout flow. Optimize for agent-initiated transactions alongside human users.
The AI Stack
Nvidia Vera Rubin Targets 35x Performance Per Megawatt
Nvidia's new Vera Rubin platform promises 35x throughput improvement per megawatt over Blackwell, with AI-powered dynamic power provisioning to eliminate wasted data center energy. Jensen Huang predicts $1 trillion in AI infrastructure orders by end of 2027.
What this means for you: Power efficiency is becoming the primary metric for AI infrastructure. When evaluating hardware and cloud providers, request per-megawatt performance metrics rather than just raw throughput. Energy costs will increasingly determine AI ROI.
Anthropic Denies Ability to Sabotage AI Tools in Military Applications
Anthropic filed a court statement denying it has any "kill switch" or backdoor access to Claude systems deployed by military customers, responding to the Trump administration's supply-chain risk accusations. A key hearing is scheduled for March 24.
What this means for you: The Anthropic-DoD dispute is establishing precedent for how AI vendor ethics policies interact with government contracts. Monitor the March 24 hearing outcome—it will signal how other vendors' policy commitments may be scrutinized.