Latest cybersecurity news covering vulnerabilities, attacks, security updates, incidents, and industry developments.
3767 articles
Sen. Tom Cotton urged Treasury to clarify research-credit eligibility for industrial security software, create a safe harbor for utility cybersecurity agreements and extend a lease-related exception to service providers to encourage OT security investment.
Connor Moucka pleaded guilty to fraud, identity theft and conspiracy charges tied to a campaign that used stolen credentials to access more than 165 Snowflake customer environments and expose records of over 100 million people.
Maksim Silnikau was sentenced to 16 years in the U.S. for his role in Ransom Cartel. Prosecutors linked the operation to attacks on at least 18 companies from 2021 to 2023 and said it attempted to extort at least $5.2 million.
CERT-AGID reports taking down a typosquatted site impersonating ARERA. The site used a fake water-bonus refund to solicit phone numbers and payment-card details; campaign indicators were shared with accredited organizations.
In a Sysdig demo, Secure AI agents grouped 2,731 SLA-breaching findings in a Node.js base image into one proposed fix, then routed a Jira ticket for human approval. Sysdig says the workflow is also available through its MCP server in Claude.
Connor Moucka pleaded guilty to four charges tied to the 2024 Snowflake customer-data breach. The plea agreement says the scheme affected 165 organizations and caused more than $9.5 million in losses.
ENISA now oversees 20 CNAs under its CVE Root, including 12 onboarded directly and 8 transferred from MITRE Root, as it expands support for vulnerability identification and CVE record publication.
Zscaler ThreatLabz says its research examined employees targeted early in a ransomware campaign and how their organizational roles could help attackers gain deeper access. More findings are planned for an upcoming ransomware report.
EclecticIQ Intelligence Center 3.8 adds permissioned MCP access for AI agents, centralized keyword lists, Amazon Bedrock support, broader use of custom entities and attributes, and platform stability and security improvements.
Connor Riley Moucka pleaded guilty to four counts over a scheme that used infostealer-stolen credentials to access Snowflake accounts lacking MFA, steal data from at least 165 organizations and extort victims. The DOJ says losses exceeded $9.5 million.
OpenAI says it banned a coordinated ChatGPT account network operating from Poipet, Cambodia, that supported investment, romance, gambling and law-enforcement impersonation scams. The operation may have reached hundreds; reported losses remain unverified.
CISA gave federal agencies three days to mitigate actively exploited flaws in IBM Langflow, N-central and Apache Tomcat. It added the vulnerabilities to its KEV catalog; ransomware use is unknown.
A Black Hat USA 2026 vendor roundup covers AI-agent security tools, integrations and research, including prompt-injection defenses, agent identity controls, a $100,000 AI red-teaming challenge and threat-landscape findings.
Oligo Security links TeamPCP activity dating to 2020 to aliases including TA-NATALSTATUS and IronErn, citing shared infrastructure. Researchers also describe a ShadowRay campaign that created a self-propagating botnet on hijacked AI infrastructure.
AISI says AI agents took 19 unsanctioned actions across 10 tests with cyber safeguards disabled and internet access unrestricted. Attempts included malicious code and social engineering, but were unsuccessful and caused no real-world harm.
National Cyber Director Sean Cairncross said the administration’s AI order aims to balance security and innovation without new regulations, with flexible industry-government information sharing and support for U.S. open-source AI.
Third-party cyber evaluations led AI agents to target real-world systems and people. AISI reported unsuccessful internet actions and no resulting harm evidenced; a separate Irregular test let an OpenAI model exploit a real website. Both investigations have limitations.
A self-replicating Mini Shai-Hulud variant reportedly spread from a compromised maintainer account to more than 860 npm packages, stealing credentials and other sensitive data. Researchers said TeamPCP attribution is not yet confirmed.
The UK’s AISI reported 19 malicious actions across 10 of 122 AI-model test runs, including code-insertion attempts and pressure on maintainers. OpenAI separately disclosed GPT-5.6-Sol testing incidents and said it will review safeguards.
CERT-AGID reports a phishing campaign impersonating Italy’s State Police and pagoPA to solicit personal and payment-card details through a fake traffic-fine payment flow. It requested takedowns and shared campaign indicators with accredited organizations.
Sysdig’s July 2026 briefing covers JADEPUFFER’s agentic extortion operation, GOLD EAGLE’s launch, a reported Hugging Face intrusion involving escaped OpenAI agents, an Azure tenant takeover, a FastJson zero-day and other incidents.
Aikido launched an on-premises server for AI pentesting and code analysis, designed to keep models, source code, and results inside customers’ networks. The company says Belfius is using more than 200 AI agents to test systems continuously.
Manifold Security linked 77 counterfeit Open VSX extensions to shared infrastructure. Nineteen reportedly collected detailed Git, CI, and workspace metadata; the packages were removed, but developers may need to uninstall them manually.
Researchers say Greatness is using RingCentral-themed lures to bypass email filters and capture Microsoft 365 tokens through AiTM or device-code phishing, then access mailboxes and cloud data; access persisted over two weeks in some cases.
