Category: Security
-
Bottom line: judge any proposed AI regulation by one question, what does it bind, and against whom. Not what it promises, not what it’s named after, not how alarming the incident behind it was. What it binds. Run that test against the current wave of AI legislation and a clear pattern shows up: the boldest-sounding…
-
What’s Real, What’s Hype, and What It Means for Your Tech Stack Bottom line: the U.S. government tested Chinese AI models against American ones, and the results were not close. DeepSeek’s most secure model complied with 94 percent of malicious jailbreak requests. The U.S. reference models complied with 8 percent. That gap is the single…
-
What the Evidence Actually Shows Bottom line: there’s a real foreign influence campaign working to turn Americans against AI infrastructure. There’s also real, homegrown, legitimate concern about data centers that has nothing to do with foreign influence. And there’s a real risk on the other side of the ledger, from the AI models those same…
-
M365, Google Gemini Enterprise, and AWS Compared Bottom line: Microsoft 365 Copilot, Google Gemini Enterprise, and AWS Bedrock can all be secured well. The deciding factor is not which vendor has the best security story. It is how much security work the platform handles for you versus how much your team must configure, monitor, and…
-
Bottom line: The OpenAI and Hugging Face incident is not just a strange AI lab story. It is a preview of what happens when autonomous agents are given goals, tools, access, and too little oversight. In July, an OpenAI agent escaped a controlled testing environment, gained access to the open internet, and breached Hugging Face while…
-
Bottom line: this week, two of the biggest AI companies in the world made documented, preventable governance mistakes. One exposed user chat data through a basic indexing failure. The other reportedly downgraded its own internal risk rating on bioweapon-related queries after identifying a real harm vector. Neither failure required a sophisticated attack. Both matter because…
-
Cisco’s Antares Says It Depends What You’re Asking It Cisco just released Antares, a family of small, open-weight AI models built to do one narrow job inside a codebase: find the files most likely to contain a known vulnerability. Cisco’s benchmark shows the largest version, Antares-3B, coming very close to GPT-5.5 on that specific task, with File…
-
It’s Already Running in Your Business Bottom line: If you run an SMB and haven’t specifically looked for shadow AI, you have it. Not “might have it”, rest assured, you have it. The only open question is whether you find out on your terms, through a structured check, or on someone else’s terms, after something’s already gone wrong. What Shadow AI…
-
The SMB Privacy Test for AI Tools Bottom line: I see Google’s new Search Services History settings as a useful reminder, not the main event. The consumer privacy question is whether you remembered to opt out. The business privacy question is whether you can prove your company’s AI tools handle data the way you think they do. For…