Amplix

Anthropic published its own account this week of three incidents in which Claude models reached real production systems during cybersecurity evaluations meant to be sealed off from the internet, a review it launched after OpenAI disclosed a related incident involving Hugging Face just eight days earlier. Three different models, across three different organizations, hit the same false premise: an environment they were told was sealed off, that wasn't. What each one did next didn't play out the same way twice, and the cause traces back to a misconfiguration at a third-party evaluation partner, not a model pursuing its own goals. See what happened in each incident and what it means for how you vet AI evaluation partners. https://hubs.li/Q04rwk6S0 #AISecurity #Cybersecurity #AIRisk #Anthropic

This week marks the end of our summer internship program at Amplix, and we want to give a huge thank you to the incredible group of interns who spent the last few months with us.   As you head back to campus for the fall semester, we're cheering you on. Good luck with your classes, your next steps, and everything ahead. We have no doubt you're going to do great things.   Aidan Cuddihy, Sohani Curtis, Aidan Connaughton, Katie Foster, Sophia Ketchen, and Julianna Bucci, thank you for choosing to spend your summer with us. The Amplix team is better for it.

The AI-takes-your-job story is a movie plot. The real story is what a job turns into, not whether it disappears. John K. Diatto, who heads channel sales for Omilia in North America, calls it commonplace iteration with technology. What changes is where the judgment sits and what the person in the role is actually deciding. "Things change, but human beings adapt. We're highly adaptable," Diatto says. New jobs get created that complement the technology as it emerges, they just look different from the ones before. #FutureOfWork #AIinTheEnterprise #ConversationalAI

Is your contact center actually behind on AI, or does it just feel that way? Amplix EVP Michael Dolloff joins Shawn Cordner on Inside Amplix to talk about where AI adoption really stands, why back-end tools like agent assist and QA are often outpacing customer-facing bots, and the vendor lock-in risk that shows up the moment token costs jump. Plus: the buy vs. build calculus, how CRM giants like Salesforce and ServiceNow are reshaping the CX stack, and Mike's advice for picking a first AI project that actually builds momentum. Watch now. 👉 https://hubs.li/Q04rbqXj0 #ContactCenterAI #CustomerExperience #AIStrategy #InsideAmplix #AmplitudeOfTech

A pilot can prove the technology works. It says nothing about whether the business is ready to run on it. Bill Fandrich, Chief Healthcare Officer at Pellera and founder of M1 Nexus, treats the real object of a pilot as the engine behind the tool, not the tool itself. That means the governance that has to evolve, the processes that need redesigning, the data strategy built to support it, the infrastructure and tech debt addressed instead of inherited. Skip that work in the name of moving fast, and the exposure isn't contained to the technology team. As Fandrich puts it, "It touches everything in your business." That's the shift from adopting a tool to building a capability, and capabilities are what scale. #AmplitudeOfTech #AIGovernance #AIStrategy #HealthcareTechnology #DigitalTransformation

IT leaders aren't bad at buying. The process they're using was built for a much simpler era, back when a technology purchase meant comparing a few vendors and negotiating price. Today the same decision touches cloud economics, cybersecurity risk, architecture, compliance, and AI readiness, and most procurement processes still haven't caught up. See the five root causes behind that gap, and the shift that closes it. ➡️ https://hubs.li/Q04r1F3N0

What does it take to give 1,500 employees room to experiment with AI without putting sensitive data at risk? Bill Fandrich, former CIO of Blue Cross Blue Shield of Michigan and now Chief Healthcare Officer at Pellera, has an answer built from forty years of IT leadership spanning mainframes, cloud, and now AI. He joins Shawn Cordner to break down his three-tier framework for AI investment, why value analysis beats a standard ROI calculation, and why vendor lock-in isn't the risk worth optimizing against right now. Listen now. 👇

The gap between a vulnerability going live and someone exploiting it has been shrinking for years, from months, to weeks, to what just happened twice in the same three-week window: seconds. An OpenAI model breached Hugging Face's servers after escaping a sandboxed test. A separate ransomware agent, JadePuffer, exploited a known, unpatched flaw and finished the job without a human at the keyboard. Different intent, same capability, confirmed twice in one month. No enterprise is fully patched. What separates the ones that hold up is how fast they can close a gap once machine-speed exploitation is the baseline. Learn more about it here: https://hubs.li/Q04qH81X0 #Cybersecurity #AgenticAI #EnterpriseSecurity #RansomwareProtection #PatchManagement

In 2024, Air Canada was held liable for a chatbot's invented fare policy. This May, a German clinic was held liable for its chatbot's false credential claims, and Google for AI search summaries that falsely linked publishers to scams.   Each one: a business caught without an answer for what its AI had said.   The best question to ask any AI vendor is why. Why did the system decide what it decided, and can they walk you through the chain that got it there?   John K. Diatto, who heads channel sales for Omilia in North America, calls this "popping the hood." Skip that question, and you're trusting a system you can't defend or fix when something goes wrong. #AIGovernance #ExplainableAI #ConversationalAI #EnterpriseAI #ResponsibleAI #AmplitudeOfTech

Four decades leading technology through mainframes, cloud, and now AI have taught Bill Fandrich the same lesson every time: it's never really been about the technology. On The Amplitude of Tech podcast, the former CIO of Blue Cross Blue Shield of Michigan and current Chief Healthcare Officer at Pellera joins Shawn Cordner to explain why so many AI investments fail to deliver real value, and what a genuine value analysis looks like next to a standard ROI calculation. Bill walks through a three-tier framework for AI investment, the responsible AI approach he built to give 1,500 employees room to safely experiment, and why vendor lock-in is the wrong thing to worry about right now. Catch the full discussion here: https://hubs.li/Q04qnHmT0 #AIGovernance #HealthcareTechnology #DigitalTransformation #ITLeadership #ResponsibleAI #AmplitudeOfTech