// AI Briefing · October 20, 2026

AI Briefing

Today’s video is on its way

The stories are ready to read now. The video usually follows within the hour.

Rendering the short

A Reuters exclusive reports that a Chinese-speaking hacker gave AI agents an instruction of fewer than 100 characters and let them break into websites at more than 100 companies, including Deere, SharkNinja and Cummins. Microsoft launched Guardian Agents and new Defender features that detect up to 60 kinds of employee-run AI agents, including Claude Code and Codex. Anthropic says it will put 5% of its shares into a new Public Benefit Fund, adding 2% more for every 50% rise in the company's value. OpenAI says its latest models are far better at writing and migrating code in legacy languages such as COBOL, Fortran and Visual Basic.

One Hacker Used AI Agents and About $5,000 to Break Into Sites at More Than 100 Companies

A Reuters exclusive reports that a Chinese-speaking hacker gave AI agents an instruction of fewer than 100 characters and let them break into websites at more than 100 companies, including Deere, SharkNinja and Cummins. Using the open-source Cairn agent platform, the agents tried more than 1,200 sites with no supervision, and NorForge researcher Snorre Fagerland estimates the token cost at roughly $50 per successful break-in. This is one of the clearest documented cases of AI agents doing nearly all the work in a real criminal hacking campaign. It suggests the cost and skill needed for mass intrusion have collapsed.

02ProductTechCrunch ↗

Microsoft Says 'Frontier Models Can't Be Trusted' and Launches Tools to Treat AI Agents Like Insider Threats

Microsoft launched Guardian Agents and new Defender features that detect up to 60 kinds of employee-run AI agents, including Claude Code and Codex. The tools also limit agents' permissions and require human approval for risky actions. Security chief Hayete Gallot said frontier models "regardless of their origin, cannot be trusted", and Satya Nadella has urged companies to separate "the supply of intelligence from the authority over it." It is a striking position from the biggest backer of OpenAI and Anthropic, and it places Microsoft as the independent overseer of the labs' agents inside companies.

03IndustryTechCrunch ↗

Anthropic Commits 5% of Its Shares to a Public Benefit Fund That Grows With Its Valuation

Anthropic says it will put 5% of its shares into a new Public Benefit Fund, adding 2% more for every 50% rise in the company's value. Ahead of an IPO that could value Anthropic around $1.5 trillion, that could be one of the largest charitable pools in corporate history. Possible uses include large gifts to nonprofits and direct payments to groups affected by AI, with job losses named as an area. Details on governance are still to come, and the move arrives as Dario Amodei calls public distrust of AI "fundamentally a crisis of trust."

04ResearchCrypto Briefing ↗

OpenAI's Newest Models Jump From 5% to 87% on COBOL Migration

OpenAI says its latest models are far better at writing and migrating code in legacy languages such as COBOL, Fortran and Visual Basic. In OpenAI's own tests, GPT-6.1 Sol scored 86.6% on COBOL migration, up from 5.4% for GPT-5.6 Sol. These languages still run core systems at banks and governments, and the programmers who know them are retiring. A jump this large could make AI-driven modernization of that old code practical, though OpenAI says humans must still review the output.

AI Briefing

One Hacker Used AI Agents and About $5,000 to Break Into Sites at More Than 100 Companies

Microsoft Says 'Frontier Models Can't Be Trusted' and Launches Tools to Treat AI Agents Like Insider Threats

Anthropic Commits 5% of Its Shares to a Public Benefit Fund That Grows With Its Valuation

OpenAI's Newest Models Jump From 5% to 87% on COBOL Migration

Previous digests