LespriCore

10/15/2026

Nearly 268,000 AI-related systems are now exposed to the public internet, according to new research from Censys.

That’s a 63% increase in just nine months.

What’s particularly interesting is how many of these tools connect AI agents to other applications, databases, and services.

An exposed system isn’t necessarily vulnerable, but it does make you wonder how many organizations know exactly what their AI tools are connected to.

We’re moving pretty quickly with AI adoption.

Hopefully, we’re keeping up with everything we’re connecting along the way.

10/15/2026

Here’s a challenging reality for vendor security assessments.

The IMF says that in more than 70% of vulnerability cases examined so far this year, software providers had almost no time to patch before exploitation.

So what does a strong vulnerability-management program look like when attackers can move that quickly?

Patch timelines still matter.

But so do exposure management, detection, containment, and the ability to respond before a fix becomes available.

Vendor assurance has to account for what happens between discovering a vulnerability and actually fixing it.

10/15/2026

Around 400 scientists are calling for an international registry of significant AI security incidents.

We actually like that part of the proposal.

Cybersecurity has benefited enormously from organizations sharing information about vulnerabilities and attacks.

AI developers are increasingly encountering failure modes that other organizations could learn from.

We probably shouldn’t need every company to independently discover the same dangerous behavior before everyone starts taking it seriously.

Around 400 scientists, including AI pioneers Geoffrey Hinton and Yoshua Bengio, on Tuesday called for an international body ​to oversee containment standards and incident reporting for advanced cyber-capable ‌AI systems.

10/15/2026

Also: Stop & Shop owner flags cyber ‘incident’; Karaoke data breach.

10/14/2026

Apparently, hackers are targeting supermarket refrigerators now.

Not to steal food.

To spoil it.

Hundreds of supermarkets in France were affected after attackers remotely manipulated refrigeration temperatures, destroying food and reportedly causing fires at some locations.

Investigators are still working to establish exactly how it happened.

We’ve connected just about everything to the internet at this point.

We’re starting to think we should have asked a few more questions before giving our refrigerators network access. 😅

French investigators were hunting the hackers who seized control of cooling units at hundreds of ​supermarkets in a cyberattack that underlined the vulnerabilities of connected systems, with no group publicly claiming responsibility.

10/14/2026

Well, that’s quite a story. 😳

The irony of OpenAI reportedly using its own AI to investigate someone allegedly spying on it is pretty remarkable.

Perhaps a glimpse into the future of how AI will be used in security investigations.

10/13/2026

10/13/2026

Apparently, hackers have discovered that poetry is a perfectly reasonable way to control malware.

Researchers found a botnet targeting AI infrastructure that extracts server addresses from words in a GitHub-hosted poem.

Change the poem, and the infected machines know where to connect next.

We have to admit, that’s pretty creative.

Unfortunately, so is the growing list of ways attackers are finding to exploit AI infrastructure.

Understand how early detection helped Black Lotus Labs uncover PoeLLM malware targeting enterprise AI systems and block its infrastructure to protect customers.

10/13/2026

Microsoft says its new AI chips can deliver roughly 40% more useful computing per megawatt than leading GPU systems.

That number might end up being more important than another benchmark improvement.

When you’re spending hundreds of billions on infrastructure, efficiency becomes a competitive advantage.

AI companies are increasingly competing on how economically they can deliver intelligence, not just how intelligent their models are.

Microsoft announced the next generation of its custom AI chip on Monday and said it plans to deploy potentially more than a million of them next year to back its internal AI services.

10/12/2026

Microsoft and NVIDIA are bringing powerful AI agents directly onto Windows PCs.

Their recently unveiled hardware will allow AI agents to handle complex tasks locally, without sending everything to the cloud.

Sounds great for privacy, right?

Well, these agents will also be much closer to our files, credentials, and applications.

We’ve spent a lot of time worrying about where AI sends our data. Now we also need to think about what it can access right on our devices.

Fun times ahead! 😄