Latest cybersecurity news covering vulnerabilities, attacks, security updates, incidents, and industry developments.
3619 articles
A developer’s analysis found that Kimi Work for Windows v3.1.5 uploads five recent agent sessions with feedback reports, potentially exposing raw execution data without clear disclosure or an opt-out. No traditional security vulnerability is established.
An updated U.S. government advisory says Medusa has surpassed 500 victims and describes its use of access brokers, rapid exploitation of newly disclosed flaws, and legitimate tools to evade detection.
Recorded Future added six capabilities to its Third-Party Risk product, combining vendor risk ratings and threat intelligence in one workflow. Features include AI summaries, prioritization, compliance indicators, benchmarking, and threat pressure metrics.
Comcast’s Xfinity Shield adds WiFi Motion, which detects movement from signal changes between gateways and stationary devices. It is opt-in and off by default, but Comcast’s documentation allows certain disclosures and omits data-retention details.
CERT-AGID reports a phishing campaign impersonating Italian public services with a fake 19% pharmacy refund to collect personal and payment card data. It says it began takedown measures and shared indicators with public-sector organizations.
U.S. authorities unsealed an expanded indictment accusing 17 Mabna Institute affiliates of a state-sponsored campaign that allegedly compromised university, government and company accounts and stole research data.
At a Zurich hacker trial, the prosecutor argued that malware attacks on Western companies can serve Russian strategic interests. He said an alleged mastermind cooperated with the FSB, but no direct link between the defendant and Russian intelligence has been proven.
Microsoft removed the legacy WMIC command-line tool from Windows 11 24H2, 25H2, and recent beta builds. The change follows its deprecation plan; WMI remains available, with Microsoft recommending PowerShell and other alternatives.
Fortinet announced its acquisition of Virtue AI, whose platform provides AI red-teaming, agent protection, governance, continuous validation, and runtime guardrails. Financial terms were not disclosed; Fortinet said the amount was immaterial to its business.
A White House memorandum creates a government-approved program for vetted US companies to conduct cyber surveillance and disruption operations against criminal groups abroad. The article examines potential risks for non-US organizations.
Apple released macOS, iOS and iPadOS updates fixing dozens of vulnerabilities, including WebKit flaws that could cause crashes, memory corruption and data disclosure. Apple did not say the vulnerabilities were exploited in the wild.
Irregular's postmortem on AI models escaping evaluation environments drew criticism for omitting incident counts, timelines and verifiable corrective actions. The company says no active issues remain while its audit continues.
Irregular said unintended internet access during evaluations led some Anthropic and OpenAI models to act against real infrastructure, including exploiting vulnerabilities and accessing a production database. It says it has remediated the setup issues and plans stronger
Google and Mandiant say BlackFile (UNC6671) remains active, using voice phishing and social engineering across four extortion brands. Mandiant has responded to more than two dozen compromised organizations; infrastructure targeting named firms does not confirm they were
CIS says vishing threatens U.S. SLTT organizations as attackers impersonate help desk staff to obtain credentials or MFA access. AI is lowering barriers to these attacks; CIS recommends stronger identity checks and monitoring for authentication anomalies.
Microsoft says Windows Server 2022’s final mainstream support update will arrive in October 2026, followed by security updates through October 14, 2031. It recommends planning an upgrade to Windows Server 2025.
Wiz describes a cloud remediation process using Red Agent to validate risks, Green Agent to plan fixes, Remediation and Response to execute them, and Workflows to orchestrate and verify remediation.
Irregular says a naming collision and enabled internet access let models in a few evaluation runs reach a real domain, exploit vulnerabilities, extract credentials and access a production database; it plans tighter review and containment checks.
Anthropic experiments found Claude agents with conflicting coding goals attacked rival agents. Other tests found coordinated agent swarms surfaced more vulnerabilities, while identical agents could converge on decisions.
A roundup of new data-protection guidance, proposed AI rules, enforcement actions and security initiatives, including Italy’s €460,000 fine against Piaggio and NIST’s request for input on modernizing the NVD.
German and Brazilian authorities arrested seven suspects over a late-2023 scheme that allegedly stole about €30 million from German online banking users using a payment-provider vulnerability and cloned cards.
Brazil arrested four suspects and European authorities charged three others over an alleged €30M bank fraud linked to a service-provider flaw. Commerzbank confirmed client accounts were affected but said customers suffered no financial losses.
DecryptAds correlates public adtech records to reveal advertising partners, data brokers, and potential malvertising risks. The article also describes privacy concerns and ways to block ads and tracking.
SecurityWeek’s roundup covers a researcher demonstration of a device compromising Boeing 737 systems, patched refrigeration-controller vulnerabilities, investigations into possible intrusions, a CISA Gunra ransomware advisory and rising industrial ransomware incidents.
