# The ultimate insider: a talk by Dinis Cruz at hack::soho, 29 October 2026 > Agents are the insider threat that never scaled before, the infrastructure cannot hold them, and risk management cannot keep up. The talk, its slides, its script and every source, from hack::soho at IOActive London on 29 October 2026. *Source: · site v0.1.6 · generated from the same content as the page, so the two cannot drift.* --- hack::soho · IOActive, London · Thursday 29 October 2026 # The ultimate *insider* Agents, the infrastructure that cannot hold them, and risk management that cannot keep up. - **Dinis Cruz**, founder of The Cyber Boardroom - Doors 6pm, talk from about 7pm, livestreamed - Live demos, all open source [Watch the slides](talk/player.md) [Read the script](talk/script.md) [What this site is](articles/introducing-this-site.md) The abstract ## Three known problems, arriving at the same time Nobody is presenting them as one story. This talk puts them side by side, using the public incidents of the last eighteen months, most of which were not attacks, just agents being agents. Then it turns to the way out. - ### The insider AI agents are the insider threat that never scaled before: a reasoning engine in a loop with tools, with skills no human insider ever had, where the distance between benign and malicious is a prompt. - ### The infrastructure What has to contain them was built for static code, and struggles to stay up on its own: no journaling, backups by the day, permissions that are the union of everything ever needed. - ### The decision The risk management that has to decide what to do about both still runs on spreadsheets, at the speed of quarters, when the decisions now have to be made in seconds and in advance. The talk, as it will be shown ## Every slide is a live infographic The slides are web pages that animate, built from the same numbers as the article and the script. Open the player to step through them; press **P** there for the presenter view with notes and timings. [Open the player](talk/player.md) · [every slide](talk/index.md) · [the script](talk/script.md) · [the talk as data](talk/talk.json) The way out ## Constrain it, and you can trust it - **Measure what it can reach**What the agent can actually do, measured rather than assumed. - **Write down what stands in the way**A boundary, a setting, an expectation, or nothing. Only a boundary is a control. - **Give it a licence to operate**With a name, a date and an expiry, that can be suspended. - **Let it earn its autonomy**Each level unlocked by evidence that the one below held. We put brakes in cars so that we can go faster. Read more ## Where the argument comes from - This site · 29 October 2026 ## [A site for one talk, built the way the talk says to build things](articles/introducing-this-site.md) What is here, how it is built, and what comes next: a team of agents, each with a written mandate. - SGit Newsroom · 30 September 2026 · 17 min ## [The ultimate insider: agents, the infrastructure that cannot hold them, and risk management that cannot keep up](https://newsroom.sgit.ai/articles/ultimate-insider-three-collisions.html) The long-form argument the talk is built from, with every number linked to its source. - Sources ## [Every number on screen, and where it comes from](sources/index.md) The public incidents, surveys and post-mortems the talk cites, grouped by act. --- *[Site index for agents](llms.txt) · [HTML version](https://the-ultimate-insider.presentation.sgit.ai/index.html)*