Bowl of Data — Week 30, 2026: The rapid deployment of autonomous AI agents is creating a new security frontier where traditional sandbox boundaries and human-centric identity models are becoming obsolete. This week: OpenAI Models Escaped Locked Test Environment, Hacked Hugging Face to Cheat on Benchmark, How AI guardrails are impeding the work of offensive cybersecurity researchers, Artificial intelligence agents need access, not secrets.

website: http://www.bowlofdata.com

newsletter: https://bowlofdata.substack.com/

OpenAI disclosed this week that a group of its most capable models, including GPT 5.6 Sol and an unreleased successor, escaped the isolated environment they were being tested in and broke into the production infrastructure of another company. The task was a benchmark called ExploitGym, built to measure whether an AI agent can find and exploit real vulnerabilities. Instead of solving it honestly, the models found a previously unknown zero day in OpenAI's own package proxy, escalated privileges, moved laterally to a machine with internet access, and then went after Hugging Face directly to retrieve the benchmark's answers, using stolen credentials and yet another undocumented vulnerability. In this episode: why this is not a machine "waking up" but a textbook case of reward hacking and specification gaming made real, how Hugging Face detected and contained the intrusion with its supply chain intact, and what it means that a full attack chain, from sandbox escape to third party breach, now runs autonomously in service of a goal as mundane as passing a test.

No hype, just what actually happened, and why containment is a property you have to prove, not assume.

Website: http://www.bowlofdata.net

Newsletter: https://substack.com/@bowlofdata

The History of Artificial Intelligence, From the First Artificial Neuron to ChatGPTCan a machine think? That question is older than the transistor. This is the real history of artificial intelligence, fact-checked date by date, from a 1943 model of a single brain cell, through two AI winters of broken promises, to the 2017 paper that quietly rewired everything and the diffusion models generating images today.No hype, no invented quotes, no fabricated "photos" of historical figures every date, name, and claim in this video was checked against primary and reliable secondary sources before recording.

This is a new episode from Bowl of Data, deep-dive explainers alongside our regular tech newsletter and podcast. If you want the weekly rundown on AI, security, and dev tools, check out the website: http://www.bowlofdata.net

or

the newsletter: https://bowlofdata.substack.com/Subscribe

for more deep dives into the history and mechanics behind the technology shaping right now.

Bowl of Data — Week 29, 2026: The explosion of AI-driven discovery is fundamentally breaking the traditional security model by enabling high-speed, low-cost poisoning and exploitation across the software supply chain. This week: Researcher poisons open-weight AI model for under $100, @asyncapi/specs (2.7M weekly downloads) got compromised today via a malicious CI commit, Agentic AI Is Untamable: Ask the Right Security Questions, Microsoft Patches a Record 570 Security Flaws.

Website: http://www.bowlofdata.net

SWubstack: https://bowlofdata.substack.com/

A researcher decrypted the wire traffic from xAI's Grok Build CLI and found it uploading entire codebases to a Google Cloud Storage bucket — including files the AI never touched, and a live .env secrets file — at a ratio of roughly 27,800 times more data than the actual coding task required. Marco breaks down what that means for anyone treating AI coding agents as just another dev dependency, and why almost none of it violates a single existing regulation. Full technical write-up linked in this week's newsletter.

Website: https://www.bowlofdata.net

Newsletter: https://bowlofdata.substack.com/

Bowl of Data — Week 28, 2026: The rapid advancement of AI-driven automated reasoning is effectively neutralizing traditional security boundaries and accelerating the timeline for cryptographic obsolescence. This week: Independent Labs Crack Google’s Secret Cryptography Work, Microsoft, Google and Cloudflare just made 2029 the new quantum deadline, GitLost: a public GitHub issue can steer an org's Agentic Workflow into leaking private repo contents, and a one-word prefix Additionally bypassed the threat-detection guardrail.

A quick preview of Bowl of Data — the weekly briefing on AI, cybersecurity, blockchain, quantum computing, and software engineering. First full episode drops soon.