https://www.zero-day.ai/Recent content on Zero Day AI LabsHugoen-ushttps://www.zero-day.ai/about/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/about/<p>Zero Day AI Labs is a small research and training group. We teach how attackers break AI systems, and how to run the ones you put in production inside a boundary that holds. We teach on code we wrote, because that is the only way to know it is true.</p>https://www.zero-day.ai/training/agents-in-production/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/training/agents-in-production/<h2 id="who-it-is-for">Who it is for</h2> <p>You have agents that work in a demo. You need them to run every day, for many users, without a person watching each one. You have to answer for what they did afterwards.</p>https://www.zero-day.ai/training/attacking-and-defending-ai-agents/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/training/attacking-and-defending-ai-agents/<h2 id="who-it-is-for">Who it is for</h2> <p>You defend, or will defend, agents that read the web, hold credentials, and run tools. You want to know what an attacker does first, and what a control must refuse for that attack to fail. You bring an agent you run today, or you use ours.</p>https://www.zero-day.ai/products/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/products/<p>We built Gibson because we kept finding the same thing on every assessment. The agent was fine. The model was fine. The thing that was missing was a line the agent could not cross, and a record of every step it took up to that line.</p>https://www.zero-day.ai/training/llm-infrastructure-on-kubernetes/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/training/llm-infrastructure-on-kubernetes/<h2 id="who-it-is-for">Who it is for</h2> <p>You run, or are about to run, models on your own hardware or your own cloud account. You know Kubernetes. You have not yet put a GPU node pool, a training scheduler, and a serving layer into one cluster and kept it up.</p>https://www.zero-day.ai/privacy/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/privacy/<p>This page says what the site holds about you. It is short because the site holds little.</p> <h2 id="the-list">The list</h2> <p>When you put an address in a form on this site, it goes to the list on beehiiv. beehiiv stores the address, the date, and which form it came from: training or consulting. We use it to send you the newsletter, the cohort dates, and the payment link. We do not sell it and we do not share it. Every mail carries an unsubscribe link, and that link removes you.</p>https://www.zero-day.ai/services/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/services/<h2 id="what-an-agentic-software-factory-is">What an agentic software factory is</h2> <p>A software factory is a place that builds, tests, and ships code the same way every time, where a reviewer can see all of it. An agentic software factory does the same for agents. Every agent has an identity. Every tool call passes a grant. Untrusted work runs in a sandbox. Every session has a budget. Every run leaves a journal a person can replay.</p>https://www.zero-day.ai/terms/Mon, 01 Jan 0001 00:00:00 +0000https://www.zero-day.ai/terms/<h2 id="the-party">The party</h2> <p>Hack The Planet LLC sells the seats and signs the assessment and build contracts. Zero Day AI Labs is the name it trades under on this site.</p> <h2 id="a-seat">A seat</h2> <p>A seat is one place in one course for one cohort. It includes every live session of that course, every lab, the recordings of every session, a tenant on the lab cluster for the length of the cohort plus one week, and the private cohort channel on Discord. A seat is for one person. A team seat is the same thing bought in a block.</p>