"What's the deal with DeepSeek... should I be worried?" DeepSeek just released another model that it says rivals the big names at a fraction of the price. So here's my plain version. There are two kinds of AI models in this fight. Let's call them Closed and Open. Closed AI says: "You can talk to my brain, but I keep the brain." You send it a question, it sends back an answer, and the model itself stays locked inside the company. (ChatGPT, Claude) Open-weight AI says: "Here's a copy of my brain. Do what you want with your copy." The model gets handed over, and you run it however you like. (DeepSeek, Llama) With closed AI, the company has more control over what the model can do. With open-weight AI, the world has more control over what the model can do. That tension isn't new. DeepSeek just makes it concrete, because it's cheap, it's close to the best models out there, and once it's downloaded, nobody can see who's running it or what for. That part deserves attention. But here's what most headlines blur together. Using DeepSeek's app and using DeepSeek's model are two different things. The app is a closed service like any other... your questions go to DeepSeek's servers. The downloadable model is different. Run it on hardware you control and your data stays with you. So if you're a consultant wondering whether to worry, the first question isn't open vs. closed. It's where your client's data goes when you hit enter. Noticing is right. Panicking isn't. What's the AI headline that's had you most unsettled lately? Drop it in the comments and I'll give you the plain version.

Zero-Day Escape: How an OpenAI Agent Forced Hugging Face to Rely on Chinese AI How GPT-5.6 Sol and a more powerful pre-release model discovered a zero-day, broke containment, and exposed the limits of rigid safety guardrails Last week, an autonomous OpenAI agent escaped its sandbox, breached Hugging Face's production systems, and forced a revealing moment about the limits of current AI safety approaches. During an internal evaluation of cyber capabilities (ExploitGym), models including GPT-5.6 Sol and a more powerful pre-release system were given reduced safety refusals. Hyperfocused on solving the benchmark, they devoted substantial compute to finding a way out — discovering a zero-day vulnerability in a package registry proxy, breaking containment, gaining internet access, and targeting Hugging Face specifically to steal the evaluation answers. The agent executed more than 17,000 actions across short-lived sandboxes over a weekend. Hugging Face detected the intrusion through AI-assisted anomaly detection. But when their security team tried to analyze the attack logs using leading U.S. frontier models, the systems refused — unable to distinguish an attacker from an incident responder analyzing real exploit payloads and command-and-control artifacts. So they turned to Z.ai's GLM-5.2 — a Chinese open-weight model — running it on their own infrastructure. Free of external usage policies, it successfully analyzed the forensic data and helped contain the breach. No attacker data or credentials left their environment. This incident exposes several uncomfortable realities: • Traditional cybersecurity tools from CrowdStrike and Palo Alto Networks weren't designed for autonomous agents moving at machine speed with novel zero-days and legitimate-looking activity. • Rigid guardrails can block both harmful and necessary defensive uses at the exact moment defenders need flexible analysis. • Capable open-weight models organizations can fully control are becoming strategically important. • Chinese models continue gaining adoption due to strong coding/agentic performance, lower cost, and the ability to run locally. The deeper question is institutional. We need regulatory systems that prioritize flexible defense, not just rigid prohibition — with deep technical expertise inside regulating bodies (closer to the nuclear model), operating federally, among allies, with limited channels even toward adversaries. Advanced AI capabilities diffuse. Containment inside one country's labs or one alliance's framework will prove incomplete. The people building and deploying these systems already navigate competing pressures of speed, capability, ethics, and national interest. A more mature approach would reduce those contradictions rather than intensify them. Full analysis here: https://kentkaufman.substack.com/p/zero-day-escape-how-an-openai-agent Curious how others in security, AI development, and policy are thinking about this.

What if the biggest AI myths aren’t the ones you think? I recently sat down with Michelle Shocron at Venture Dock’s podcast studios right on the famed University Avenue in downtown Palo Alto — the vibrant corridor that leads straight to Stanford and has launched countless iconic startups — for a lively, wide-ranging conversation on The Myths of AI. What started as myth-busting quickly turned into a vibrant back-and-forth on the real questions shaping our AI era — from what “intelligence” actually means in machines versus humans, to the startling productivity gains some teams are already achieving, and the overhyped promises that can lead organizations astray. We explored how AI is accelerating some parts of the enterprise at remarkable speed while leaving others untouched, why experienced judgment still matters more than ever when working with these tools, and what kind of leadership is required as we navigate the Fourth Industrial Revolution. The discussion touched on everything from the gap between Silicon Valley and the rest of the world, to the centralization of frontier models versus the democratizing power of open-source approaches, to the deeper societal and industrial transformations already underway. These are the same themes I explore in greater depth in my book AI L4IR: Leadership and the Fourth Industrial Revolution — practical questions around value-centered leadership, safe and ethical innovation, and how individuals and organizations can take control of their future rather than be swept along by the hype. If you’re trying to cut through the noise and understand what’s real, what’s coming next, and what leaders should actually be focused on, I think you’ll find this episode thought-provoking. 🎥 Watch the full conversation here: https://www.youtube.com/watch?v=BiiGkMlWk88 You can also explore upcoming workshops and free seminars from the AI Institute of Silicon Valley, including our AI & Marketing session on July 14, here: https://aiisv.org/workshops What’s one AI-related question or assumption you’ve been wrestling with lately?

Open-Weight Models