AI leaks, billion dollar labs, and the new retail battlefield...

Today’s dispatch spans from corporate secrets quietly bleeding into chatbots to a pharmaceutical titan going full Iron Man lab with Nvidia. We’ve also got one of the biggest retailers in the U.S. doubling down on AI shopping agents… and finally, a CEO making a near total human swap out in favor of algorithms. Let’s get into it.


Shadow AI… when your employees go rogue with ChatGPT

Nearly half of employees say they’ve uploaded sensitive company data into AI chats like ChatGPT… and most of it’s happening totally off the radar.

That’s according to a new Stack Cybersecurity report sounding the alarm on “Shadow AI.” In absence of clear enterprise policies, guardrails, or even approved tools, employees are freelancing with generative AI… feeding it confidential info via personal devices well outside IT’s reach. The result? Forty eight percent of workers admit they’ve entered sensitive company or customer data into AI tools, and 44% readily admit what they’re doing violates company policy.

The reasons aren’t malicious… they’re practical. Employees are using AI to get their jobs done faster. But without sanctioned systems, they’re innovating their own pipelines… often through browser extensions and non-enterprise versions of tools like ChatGPT and Claude. And when compliance and privacy teams aren’t looped in, the risk to trade secrets, client data, and regulatory exposure multiplies fast.

Stack is throwing down a lifeline with their new AI Hub… a free resource suite including an AI readiness evaluation and a customizable policy template (plus a custom ChatGPT instance trained to help companies get AI ready). The takeaway for leaders? If you’re not proactively managing AI use across your organization, your team already is … just without you.

Lilly teams with Nvidia on $1B AI lab to accelerate drug discovery

While most startups are tightening budgets, Eli Lilly and Nvidia are betting big… a billion dollars over five years… on a new AI co-innovation lab to radically speed up drug development.

Announced at the J.P. Morgan Healthcare Conference, the partnership is a heavyweight combination of Lilly’s R&D muscle and Nvidia’s AI prowess. The incoming lab, set to open in the Bay Area by Q1’s end, will use multimodal foundation models, physical AI, and robotic platforms to build what NVIDIA’s Kimberly Powell calls a “closed loop discovery” system. That means a tight, iterative pipeline where AI models are trained on real world lab data, optimizing everything from target identification to clinical trial readiness.

Lilly’s executive team signals a shift in mindset here: AI isn’t just boosting productivity… it’s now a scientific collaborator. Tuning AI models with Schrödinger's LiveDesign software, building supercomputers for pharmaceutical use, and running open platforms like TuneLab and BioNeMo show they’re dead serious about making AI core to discovery, not auxiliary. Biotech founders and VCs, take note… this is what the future of verticalized AI looks like… massive proprietary data + purpose built compute + deep integrations.

The AI shopping agent wars officially begin

Retail’s next battleground won’t be fought over low prices or faster shipping… it’s going to be decided by who has the smartest AI agent helping you buy drywall anchors at 10 PM while you’re on the couch.

Google Cloud and The Home Depot just escalated it. At NRF 2026, they unveiled a slate of agentic AI tools that turn retail from “search and scroll” into “just tell the AI what you want.” The updated Magic Apron Assistant now offers multimodal support (read: show it a picture of your broken sink) alongside AI generated product lists, local inventory visibility, aisle level wayfinding, and even route optimized delivery logistics. Piped in through Gemini models across SMS, chat, voice, and search… this move embeds AI across every CX touchpoint inside and out of stores.

Importantly, they’re not stopping at the customer. The Home Depot is also using Gemini Enterprise for internal workflow automation including project bottleneck prediction and generative creative tasks… showing how agentic systems can compress operations from backend to storefront.

For founders in retail tech… UX expectations just leaped forward. The magic is no longer just sleek design or same day delivery… it’s personalized, persistent AI agents that remember where you are (and what shelving unit you bought last spring).

When AI takes over and calls it kindness

The most radical AI reorg of the week comes from IgniteTech, where CEO Eric Vaughan laid off nearly 80% of the company to replace them with AI systems… publicly framing the decision not as ruthless, but “kinder.”

The shift was abrupt. In Vaughan’s view, internal resistance to the company's aggressive AI pivot was untenable. Employees clinging to legacy processes or worried about job displacement were either reassigned or removed. The unapologetic justification? Those who embrace AI should stay. Those who don’t, shouldn’t. Vaughan argues that delaying the transition is more damaging than a clean, if brutal, reset.

There's philosophical debate here (and let’s be clear… PR blowback too). But strategic implications loom even larger. AI native companies are not only becoming structurally leaner… they’re embracing a zero redundancy model. This isn’t automation as augmentation. It’s full replacement, from engineering to marketing, operating under the bet that LLMs + lean ops will outperform traditional headcounts.

Founders and operators thinking of similar shifts will need to answer tough cultural and ethical questions… especially when vision outpaces workforce readiness.

Looking ahead

AI isn’t just digitizing industries anymore… it’s remaking their infrastructure. Whether it’s quiet compliance threats inside your org or large scale bets on multimodal biotech labs, the pattern is clear… game changing capabilities are raising equally game changing risks and expectations.

Governance is catching up too slowly. Incumbents are moving faster than they’re getting credit for. And the opportunity for startups? Sitting right at the edge of these gaps… whether in security, tooling, or vertical integrations.

More tomorrow.