Security | Threat Detection | Cyberattacks | DevSecOps | Compliance

AI desktop assistants and coding tools need credentials to reach external services, and many of them store those credentials as plaintext JSON at predictable paths in the user's home directory. This research covers how credential storage works across 14 popular AI tools, where OS keychain integration is present or missing, and eight attack scenarios that turn that exposure into real risk, from malware-based theft to remote session hijacking to supply-chain compromise via MCP servers.

Mini Shai-Hulud is back. Like I said before, we were yet to see the full scale of the attack. The npm campaign we covered in April, when it targeted SAP packages, has now turned into a much larger compromise. Our Malware Team detected 373 malicious package-version entries across 169 npm package names. The basic goal is still the same: steal credentials from developer machines and CI/CD runners, then use those credentials to reach more packages. What changed is the scale and the release path.

Using Cursor, GitHub Copilot, Claude Code, Codex, or another coding agent means giving software access to more than your code. It can also see the credentials available in your workspace, shell, config files, and development environment.

Deception and Claude Mythos is no longer just a cybersecurity conversation. It’s a paradigm change in how organizations need to consider cyber resilience in the era of AI. Enterprises are facing a new threat landscape in which attackers can find exploits quicker, conduct enterprise-wide reconnaissance with low-level automation, and exploit enterprise assets with unprecedented accuracy on an enterprise-wide scale.

AI-powered attackers are faster and more systematic than ever. But they still trust what they see. Deception technology controls what they see. 87% of security leaders say AI-related vulnerabilities grew faster than any other risk in 2025 44% year-over-year rise in exploitation of public-facing applications in 2025 300K+ AI platform credentials exposed via infostealer malware on dark web in 2025.

As I write this, we are now almost 2.5 years post acquisition of VMware by Broadcom. In that time many things have changed. The partner program has shrunk and shrunk again. The portfolio is much smaller as non-core product lines such as VDI (now Omnissa) and Carbon Black have been divested. However, none of these changes have been as impactful as integrating all remaining capabilities into the evolved VMware Cloud Foundation (VCF).

The Tines Voice of Security 2026 report found that security professionals spend 44% of their time on manual, repetitive work. A workflow engine is the software built to take that operational drag off people, deciding what happens next based on events, rules, and state. The category is shifting. The workflow engine used to live inside one system, running a narrow set of backend steps.

Most security programs have more visibility than ever. Dashboards are full. Alerts are firing. And incidents are still happening. That contradiction is not a coincidence. It reflects something most security vendors have quietly avoided saying out loud: Visibility and control are not the same thing, and for a long time, the industry has been selling one while calling it the other.