digit.fyi · 7 min
Cato Networks showed an AI agent stack achieving Domain Admin in 40 minutes, recon to exfiltration, fully automated. It's a proof of concept, but the speed is the point.
AI roundupSunday 19 July 2026
Today's stories are split between worrying about what AI agents can already do and watching Big Tech muscle into new verticals. A controlled experiment showed an AI-driven attack chain compromising an entire domain in under an hour, while six major tech firms made legal AI moves in four months. The gap between capability and governance keeps widening.
In today's edition: Agentic threats, Big Tech's legal play and Bias and governance, plus quick bites.
The most alarming demonstration of autonomous AI capability this week comes from a controlled red-team experiment.
digit.fyi · 7 min
Cato Networks showed an AI agent stack achieving Domain Admin in 40 minutes, recon to exfiltration, fully automated. It's a proof of concept, but the speed is the point.
While one agent was breaking into networks, six big tech firms were breaking into the legal industry.
news.bloomberglaw.com · 5 min
Microsoft, Anthropic, OpenAI, Google, SpaceXAI, and Amazon all made legal AI moves in four months. The gold rush for law firm contracts is on, and Big Tech is arriving with purpose-built tools.
Meanwhile, the real-world consequences of deployed AI systems continue to surface in courts and schools.

sea.mashable.com
A class action lawsuit in Ontario alleges an AI risk-assessment tool disproportionately places Black prisoners in maximum security. The province reportedly knew about the racial skew. This is the kind of story that keeps regulators up at night.

sea.mashable.com
Education advocates are urging schools to cut ties with Google over AI Search's unreliability. The irony: the generation taught to distrust Wikipedia now has to distrust Google itself.
<em>British Journal of Educational Technology</em> | BERA Journal | Wiley Online Librarydoi.org
AI Gateway - Vercelvercel.com
Feels like a watershed moment for advancing mathematics. Scientific and medical advances which can really improve people’s lives feeling very close now.erdosproblems.com
Have Chinese AI Models Caught Up to the US Frontier? I have spent the last 2 days writing this article. It should settle the debate once and for all.open.substack.com
How UK Businesses Can Improve AI Governance Without Slowing Innovationprivacyneedle.com
MongrelDB-Hermes (Memory & Context)hermesatlas.com
Claude iOS Bottom Navigation Bar: Anthropic Tests New Mobile Navigationprogressiverobot.com
What a Deepfake Fraud Incident Teaches Companies About Data Protectionprivacyneedle.com
Netflix pays $587 Million for Ben Affleck’s AI startuphappymag.tv
MidJourney Video AI Free: Create Stunning Videos in 2026resource.digen.ai
Chinese Smartphone Companies Rework Devices Around AIpymnts.com