What Happened
Microsoft observed attackers accessing cloud-hosted ScreenConnect instances used by managed service providers beginning November 11, 2026. They installed unauthorized remote-access clients and SCKit on managed endpoints. SCKit persists through registry entries and scheduled tasks, adds Microsoft Defender Antivirus exclusions, seeks browser authentication cookies, and attempts to spread to technicians when they connect to…
What Happened
Microsoft researchers showed that poisoned fine-tuning data and attacker-controlled model checkpoints can teach coding agents to make unsafe tool calls in response to otherwise benign requests. In one proof of concept, 90 poisoned examples among roughly 9,000 conversations produced unsafe calls in about 78% of trials on two familiar agent setups, but only…
What Happened
Recent research illustrates two connected risks: an AI agent may be manipulated before deployment, and the systems hosting it may be compromised through conventional vulnerabilities.
Microsoft Research demonstrated that changing fewer than 10 model neurons could cause an agent to pass evaluations yet follow an attacker’s command when triggered. The researchers have not…
What Happened
Microsoft Research showed that editing fewer than 10 neurons could make an agent model pass evaluations yet execute an attacker’s command when a trigger string appeared. The attack required only inference-level compute. Microsoft said it had not observed threat actors weaponizing the technique. [1]
Separately, Microsoft’s FORGE Lab used agentic tools to find…
What Happened
Microsoft’s FORGE Lab used Hyenas, a research version of MDASH, to identify and report memory-safety defects in Rust’s standard libraries. Some defects could cause undefined behavior through entirely safe Rust code, because the unsafe implementation beneath a safe API relied on assumptions that did not always hold [1].
The findings involved panicking callbacks…
What Happened
Microsoft reports that Cambodia-based Storm-3358 began using the agentic AI framework ARTEX in October 2026 to target financial organizations, primarily in Southeast Asia. The group scanned and exploited internet-facing systems, then used WISP, vshell and AdaptixC2 remote-access tools for persistence and lateral movement. ARTEX can accelerate web exploitation, credential theft, cloud identity abuse…
What Happened
Two recent analyses highlight attacks that can bypass familiar security assumptions. A review of Binance’s October 2026 incident reports that attackers used social engineering, an internal AI tool, stolen credentials and access to a wallet-signing service to obtain valid signatures without stealing private keys. An EIP-7702 delegation then gave the attacker persistent wallet…
What Happened
Microsoft observed threat actors using the agentic AI framework ARTEX to find secrets in public GitHub repositories. In one case, an agent progressed from finding client secrets to exfiltrating Azure Key Vault data in five minutes. Stolen application credentials were also used to sign in as Microsoft Entra workload identities and search Microsoft…
What Happened
Microsoft Threat Intelligence identified fake downloads for software, games, plugins, and utilities that use trojanized Inno Setup-style installers to deliver multiple payloads in one session. Observed branches installed proxyware, compute software, or other utilities. One proxyware branch created SYSTEM scheduled tasks for startup, watchdog, and updater functions. Because delivery domains rotate and one…
What Happened
Microsoft’s analysis found that 2,056 of roughly 383,000 CVEs disclosed from 1999 through September 2026 were linked to exploitation in the wild—about 0.5%. Vulnerabilities enabling system takeover, including remote code execution and privilege escalation, accounted for 78% of exploited CVEs. In 2025, 29% of observed exploits launched within a day of CVE publication…
Malicious AI Coding Extensions Can Leave SSH Backdoors After Removal: How to Detect and Contain Them
What Happened
Tampered versions 2.1.312 and 2.1.313 of the Claude Code-branded anthropic.claude-code extension were served through the Cursor Extension Marketplace on October 29, 2026, from approximately 2:27 AM to 1:35 PM PDT. Available evidence points to that marketplace alone; the Microsoft Visual Studio Marketplace, npm, and Anthropic systems were not affected [1].
The injected JavaScript…
What Happened
Two recent intrusion investigations began with attackers impersonating IT staff in Microsoft Teams. They persuaded users to grant Quick Assist access, then used staged payloads and portable runtimes to inspect the environment and target Active Directory Certificate Services (AD CS). In one case, a vulnerable certificate template enabled issuance of certificates for privileged…