No confirmed exploitation is the important qualifier here, and it matters. PhantomRPC is a research disclosure at this stage, not an active threat in the way that RedSun and UnDefend are. That said, the combination of factors that makes it worth watching is clear. The research is fully public, the tools are on GitHub, the flaw affects every version of Windows, there is no patch, and there is no patch timeline. That is the same setup we saw with Defender RedSun and UnDefend before those moved from disclosed to actively exploited. The five exploitation paths described in the research vary in how difficult they are to execute. Some require an attacker to already have low-privilege code execution. PhantomRPC alone does not give initial access but it gives a reliable privilege escalation path once initial access is achieved through any other means. In that context it pairs with the kind of initial access that has appeared repeatedly in this brief: compromised developer credentials, RMM platform exploits, Teams social engineering, OAuth token theft. Any of those initial access paths followed by PhantomRPC escalation produces a SYSTEM-level compromise. The practical action today is monitoring rather than emergency patching since there is nothing to patch. Kaspersky's released tools allow auditing environments for vulnerable RPC call patterns. Watch for a Microsoft response and keep this item in view.

This is Issue 13's Bitwarden supply chain story at a different layer of the same attack surface. Yesterday it was a backdoored CLI package on npm. Today it is weaponized VS Code extensions on Open VSX. Both target the developer environment specifically because a compromised developer machine is not just one endpoint. It is a gateway to every repository, pipeline, and production system that developer touches. The targeting of AI coding tool extensions is deliberate, and it points to a broader risk that most practitioners have not fully internalized. AI coding tools have compressed the time between "I need a package that does X" and "I have it installed and running." A developer working with Claude Code, Copilot, or Cursor gets a package suggestion and a ready-to-run install command in the same response. The friction of checking what that package actually is before running it feels like it slows things down. That friction gap is exactly what attackers exploit. This dynamic extends beyond VS Code extensions. Python packages installed via pip carry the same risk. PyPI has been targeted repeatedly with typosquatting attacks where a package name one character off from a legitimate one delivers a credential stealer. A developer who asks an AI tool to help them parse JSON, make an HTTP request, or interact with an API and then runs the suggested pip install command without checking the package is trusting a chain that includes the AI model's training data, PyPI's moderation, and the package maintainer's account security. Any link in that chain can be compromised. The practical habit worth building is treating any unfamiliar package name, whether suggested by an AI tool or found in documentation, as something worth a 30-second check on PyPI or npm before installing. Download count, maintainer history, and publication date catch most malicious packages before they land.

The category of data exposed in a security monitoring breach is different from a typical retail or financial breach and deserves specific attention. ADT's business is knowing where customers live, what their alarm codes and access schedules are, which entry points are monitored, and when properties are typically occupied or vacant. That operational intelligence has value beyond identity theft. Physical security researchers and law enforcement have documented cases where breach data from security monitoring companies was used to plan burglaries, circumvent alarm systems, and time physical access to properties when they were known to be unoccupied. For businesses that use ADT for commercial monitoring, the concern extends to understanding which employees have alarm codes, when buildings are typically secured, and what the monitored access points are. A breach notification in this category should prompt a review of physical access codes and procedures, not just a credential rotation exercise. This is the third time ADT has disclosed a breach in three years, which raises a different set of questions for any customer about the maturity of their security operations. The immediate practical step is to rotate alarm codes and access credentials for any property monitored by ADT regardless of whether you receive a direct breach notification.