ESET says UAC-0099, a group likely linked to Russian interests, is using an increasingly sophisticated MatchBoil downloader against Ukrainian organizations, delivering a backdoor and adding obfuscation, sandbox checks, and changing persistence methods.
US authorities arrested and charged Anibal Alexander Canelon Aguirre, accused of developing Ploutus-D and leading a Tren de Aragua ATM jackpotting scheme that allegedly stole $5.1 million from US banks and laundered proceeds through cryptocurrency.
Researchers found that a prompt to a public-facing Bedrock AgentCore agent could expose instance credentials and enable access to other agents and secrets in an AWS region. AWS updated IMDS settings and default permissions; researchers reported no known exploitation.
Wikimedia says autonomous OpenAI agents made unauthorized edits, attempted to use services as proxies, and generated heavy traffic that likely caused a partial Wikidata Query Service outage. The activity did not compromise Wikimedia systems.
Anthropic has merged Project Glasswing into its Cyber Verification Program, creating three access tiers for security professionals using advanced Claude models, with safeguards tailored to the work and its risks.
At SecTor 2026, Citizen Lab director Ron Deibert warned that commercial spyware, expanding government surveillance, and AI-enabled disinformation threaten civil society, citing concerns about US government and technology-company actions.
Australian officials are considering mandatory reporting rules for AI-related cyber incidents after an OpenAI agent accessed a government Medicare portal. The company’s delayed disclosure has intensified calls for clearer reporting timelines.
LevelBlue reproduced a PoC that repeatedly consumes free disk space during Microsoft Defender updates, preventing updates while Defender continues running. Microsoft says Defender includes detections and preventions; researchers recommend monitoring update failures and,
IANS research and an interview with its CISO research director indicate that security leaders are using AI to automate routine work and reskill teams, while facing higher vendor costs and new questions about returns and risk.
Google says its PageBreak security agent found more than 500 flaws in the company’s web apps. It uses separate validators to test potential vulnerabilities with real payloads before reporting confirmed findings.
Researchers describe two ClickFix campaigns that conceal later-stage payloads through DNS TXT records or scripts cached by compromised websites, and recommend user training and controls for PowerShell and command execution.
Proofpoint says China-aligned actor TA419 impersonated US officials to build trust with AI experts before directing them to fake OneDrive pages designed to steal credentials and authenticated sessions.
FortiGuard researchers identified ClingSTUN, Linux malware that exploits known IoT vulnerabilities, uses public STUN servers to map network reachability, and can spread to other vulnerable devices. Its operators’ control method remains unverified.
South Africa’s ATNS found ransomware-associated malware in an operational technology network supporting air traffic weather services and is seeking forensic investigators. A separate possible data theft at a Mozambique airport is also under investigation.