PaperCut attackers appear to be moving beyond probing into active compromise, as multiple related vulnerabilities are increasingly associated with real-world intrusions. CISA added PaperCut NG/MF issues (including CVE-2026-82078 and CVE-2026-81578) to its Known Exploited Vulnerabilities (KEV) catalog, signaling that exposed instances need urgent remediation. For defenders, the headline takeaway is timing: when printing platforms are exposed to the internet, delays in patching can quickly translate into installed access.
Source: SecurityWeek
PaperCut servers: attackers planted legitimate remote access tools after exploiting print systems
PaperCut reports that the ongoing threat targeting internet-facing Application Servers includes covert installation of legitimate remote access software on compromised hosts. The vendor previously advised restricting web access to trusted IPs only, underscoring how attackers are leveraging real pre-auth weaknesses to establish durable access. Organizations using PaperCut should treat this as a “patch-and-verify” incident: review logs, search for remote access tooling, and confirm that any temporary exposure has been fully removed.
Source: Help Net Security
Anthropic locks out Claude users after infostealers hijack sessions and drain paid usage
Anthropic says infostealer malware campaigns have compromised active Claude login sessions, allowing attackers to bypass normal account protections and consume paid usage without stealing passwords. Users are being logged out and payment data removed as a containment step, and Anthropic lists multiple malware families observed in these incidents. The key lesson for security teams is that session theft is functionally equivalent to credential compromise—monitoring and detection must extend to authenticated session integrity, not just login events.
Source: Help Net Security
McKesson confirms cyber incident tied to ShinyHunters’ claim of massive patient data theft
McKesson disclosed a cybersecurity incident involving unauthorized access through a third-party application, while attackers (via ShinyHunters) claim theft of 284 million patient records. Multiple reports emphasize early-stage investigation language and the likelihood of service degradation as systems stabilize and forensic work continues. For healthcare operators and their partners, the risk signal is supply-chain adjacency: third-party access paths can be the shortest route from an initial foothold to large-scale data exposure.
Source: Help Net Security
Threat actors spoof “AI crawlers” to find exposed credentials and config files
Researchers warn that attackers disguise automated scanning as traffic from major AI platforms (e.g., “ClaudeBot,” “Claude”/“OpenAI” style identifiers) to locate exposed credentials and sensitive configuration files. Because user agents can be trivially spoofed, defenses should not trust crawler-like traffic claims at face value. Expect more credential-harvesting reconnaissance to blend into “legitimate” machine traffic patterns, so web monitoring and secret-exposure detection need tighter validation than simple header checks.
Source: Help Net Security
“The Coding-Agent Trap”: a “free” LLM endpoint can be the adversary
SANS ISC highlights a practical scenario where an internet-exposed inference honeypot was discovered, relabeled with attractive model names, and used to conduct coding-agent activity. Instead of merely observing requests, the adversary leveraged the setup to gather rich operational details—history, filesystem output, local tool manifests, and agent context. The takeaway is operational: if your agent or tooling trusts an external “endpoint” promising free capability, you may be handing the attacker your environment and workflow data.
Source: SANS ISC
You May Also Be Interested In...
CISA adds additional PaperCut NG/MF flaws to the KEV catalog
Critical Ruby on Rails file read flaw reportedly targets secrets and remote code execution
Debian developers reject an LLM ban; disclosure becomes voluntary