ChatGPT Can Read Every Tab You Have Open. Including the Client’s.
ChatGPT’s Chrome extension can now read across your open tabs and act on them. If you manage multiple clients in one browser, that is a confidentiality problem.
Cybersecurity & Privacy
Security and privacy news translated into plain-English steps for small teams: AI browser and agent risks, WordPress and plugin patching, account access, and client confidentiality. Each story points to the practical controls worth checking first.
Assume Your AI Is Compromised. Then Build Like It.
Microsoft’s CEO says treat every AI model as already compromised and contain it from the start. Four controls, translated for a business running AI agents.
Meta’s Scraping Clause Is Gone. Its Crawlers Are Not.
Meta pulled the clause letting it use crawled site content for AI training. Nothing about its crawlers changed. Here is what to actually lock down.
AI Now Hunts Bugs in the Code Your Website Runs On
Anthropic is scanning open-source projects for vulnerabilities with its strongest models, free and without human review. Your patch window just got shorter.
AI Agents Hit One UN Site 16,000 Times. Check What They’re Doing to Yours.
A researcher logged 16,000+ requests from OpenAI agents to one UN statistics site in three months. Your hosting bill and your analytics are on the same list.
OpenAI’s Own Agents Leaked 53 User Images. Audit What Your Team Uploads.
OpenAI says its agents posted 53 user images to public hosting sites and it cannot identify who was affected. Here is the upload policy your business needs.
Google Froze Its Bug Bounty Over AI Slop. Your Contact Form Is Next.
Google paused its open source bug bounty because AI-generated reports overwhelmed human review. The same intake math is quietly breaking business lead forms.
Your AI Agent Will Break the Rules to Finish the Job
Agents don’t need bad intentions to cause an expensive week. They need a goal, your credentials, and a system with a soft spot. What every small business should lock down first.
An AI Found a Zoom Takeover Bug in Under 20 Prompts. Your Patch Habits Are the Real Story.
Three Zoom annotation flaws let another person on your call reach your machine. Zoom patched fast. The gap that should worry a small business is the seven weeks between the fix shipping and anyone installing it.
Your AI Agent Logs Are Leaking Passwords in Plain Sight
Researchers decoded 315,320 “encrypted” AI thinking blocks from public agent logs and pulled out live API keys and passwords. The geopolitics got the headlines. The part you can act on didn’t.