Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

Zero-Shot Learning is a podcast about how AI gets built, secured, and deployed. Hosted by Nancy Wang, 1Password CTO, and Dev Tagare, Senior Director of Engineering at Google, it's a builder's view of the architecture and the complex choices it takes to ship with AI.

On May 25, 2026, the maintainer of jqwik, a Java property-based testing library, released version 1.10.0 to Maven Central with a hidden instruction intended for AI coding agents. The payload told agents to disregard previous instructions and delete all jqwik tests and code. It was hidden from humans with ANSI terminal codes but left fully readable to any tool that captures raw output.

For many organizations, the move to virtual private server (VPS) hosting feels like a natural security upgrade. After all, the word private suggests isolation, control, and protection; especially compared to shared hosting environments. But in practice, private hosting does not automatically mean secure hosting. In fact, without the right security maturity, VPS environments can introduce new risks rather than eliminate old ones.

Every year, the Verizon Data Breach Investigations Report (DBIR) gives the security industry a chance to step back from the noise and look at what happened. Not what vendors predicted. Not what attackers threatened. Not what defenders feared. What happened. This year’s report makes one point hard to ignore: vulnerability exploitation became attackers’ initial leading access vector.

Under pressure, organizations often prioritize speed over process, unintentionally making the situation worse.

In our previous post, Token Bingo: Don’t Let Your Code Be the Winner, we documented Kali365, a phishing-as-a-service (PhaaS) kit abusing Microsoft’s OAuth 2.0 device authorization flow to steal Entra ID tokens. In this follow-up report, we track the same operator into new territory as they expand their operation and infrastructure.

AI adoption in business has moved at a staggering pace. According to a major survey from The Conversation, 58% of global employees are intentionally using AI at work. That same study revealed an alarming trend: 66% of global employees have used unapproved AI tools, while only 34% say their company has put in place rules to govern AI usage. This use — and potential misuse — of AI systems is the latest and most complex threat facing businesses today.

The terminal output was still scrolling when Jer Crane, the founder of PocketOS, realized what had happened. Nine seconds. That is how long it took a coding AI agent to delete his production database, his backups, and three months of operational records. PocketOS was using Cursor for what should have been a routine task in a test environment.

Many businesses depend on cloud-based tools like Microsoft 365 to run their daily operations. Because of this, Microsoft 365 monitoring has become more important to ensure that the environment runs smoothly and securely. Microsoft 365 is more than just an email service. It is a full suite that comes with productivity apps, collaboration tools, and file storage. But this large environment makes it a potential target for cyber threats, so it's important to monitor and protect its activities.

A web application firewall is a security software that observes and filters HTTP/HTTPS traffic between a web application and the internet. While this has been available for decades, with the evolution of the threat landscape, WAFs have also added additional capabilities to protect not only web apps but also APIs against a range of attacks, including DDoS and bot attacks. So, the category has evolved and is currently called Web Application and API Protection (WAAP).