Blog | Greg Brown

23 Jul 2026

Sandbox escape: guardrails are policy, containment is architecture (OpenAI / Hugging Face)

Imagine someone broke into your home, searched through your belongings looking for information about you, and stopped only when you caught them. That is close to what happened during one of OpenAI's recent cyber capability evaluations. Two models were being tested on an internal benchmark. Their cyb

AI governance SME strategy AI cybersecurity agentic AI risk

Read more →

17 Jul 2026

Inherited robustness: you cannot run the lab's test, and the evidence expires (GPT-Red). Model choice as a security control. News-pegged.

Most organisations believe they have a rough sense of how exposed their AI systems are. Something OpenAI published this week shows how much of that judgement rests on evidence they cannot reproduce. It has built GPT-Red, an internal automated attacker trained specifically to break AI systems. By the

vendor dependency SME strategy AI cybersecurity prompt injection agentic AI risk

Read more →

16 Jul 2026

The enablement gap: labs now sell the build, not the model (Ode + Deployment Company). "You can buy the build, not the fluency." News-pegged.

Anthropic’s $1.5bn implementation venture now has a name: Ode. The more interesting fact is that it exists at all. Within a week in May, Anthropic and OpenAI both announced separate businesses dedicated not to building better models, but to making those models work inside companies. Anthropic has Od

AI adoption SME strategy enablement forward deployed engineers vendor strategy

Read more →

09 Jul 2026

The measurement blind spot: assisted output hides the capability gap (Brown 96->48). "What is your in-person score?" News-pegged.

An economics professor at Brown University recently ran, almost by accident, one of the most useful experiments of the AI era. He set his class a take-home midterm exam. The average came back at 96 out of 100, the highest he had ever seen. Suspicious rather than flattered, he set the next exam in pe

AI adoption SME strategy capability gap skills measurement

Read more →

02 Jul 2026

Digital sovereignty becomes operational risk: SCOTUS FTC ruling meets EU Tech Sovereignty Package (news-pegged). Image: EU Commission slide.

Digital sovereignty is easy to admire and easier to defer. It is a phrase European leaders reach for at conferences and forget on the way back to their offices, where their data still sits on a US cloud, and the AI runs on American models. The arrangement that made the deferral comfortable is the EU

SME strategy digital sovereignty data protection EU-US data transfers vendor lock-in

Read more →

30 Jun 2026

AI security's human bottleneck: finding got cheap, fixing didn't (Patch the Planet); runs before jagged-frontier post

The reassuring version of AI in cybersecurity runs roughly like this: the machines find the flaws, the machines write the fixes, and software slowly gets safer on its own. This is the premise on which GPT-Cyber works. It is half right, but the missing half is the one that matters. What AI has actual

vendor dependency SME strategy AI cybersecurity open source supply chain risk

Read more →

23 Jun 2026

AI as utility: settled product, unsettled economics (OpenAI losses + SpaceX compute landlord)

We treat AI like we treat electricity: something that simply works, will keep working, and will hopefully cost roughly what it does today. This cannot continue indefinitely. Consider the commercial layer first. According to leaked financials, OpenAI, the firm behind ChatGPT, took in around thirteen

AI economics compute platform risk vendor dependency SME strategy

Read more →