OpenAI is introducing an invisible watermarking technology called textGrain to text generated by ChatGPT and Codex in the European Union. This watermark is not visible to users but creates a detectable statistical pattern in the AI-generated text. The watermarking is opt-in for API developers globally and currently limited to the EU by default. Editing the watermarked text, such as replacing words with synonyms, significantly reduces detection accuracy. The watermark does not reveal user identity or the extent of human editing. OpenAI states that watermarking does not meaningfully affect the quality of the generated text.
The SelectorsHub Chrome extension, used by approximately 400,000 users, has been found to open advertisement tabs selected by its server without any user interaction. The extension fetches ads from a server that can change the URLs dynamically without extension updates. It also collects cookies broadly and sends user selectors to a domain that currently redirects to a gambling site. This behavior resembles adware and poses privacy and security concerns, especially for users in sensitive environments.
This content is an invitation to a Reddit Ask Me Anything (AMA) session hosted by IBM Security experts Jeff Crume and Guido Crucq. They discuss cybersecurity trends and findings from the Cost of a Data Breach Report, which analyzes real-world data breaches. No specific vulnerability, exploit, or active breach details are provided.
Italy's Data Protection Authority fined IQVIA €7 million ($7.8 million) for inadequate anonymization of health data, which risked exposing and de-anonymizing roughly one million patients. The company aggregated data from 800 general practitioners, using unique codes that could be linked back to individuals when combined with detailed personal and medical information. Additionally, IQVIA processed data without proper legal basis or patient notification, violating GDPR, and failed to establish data retention periods, with records dating back to 2001. A subset of 3,300 patients had personally identifiable information included. The authority ordered IQVIA to comply within 120 days. IQVIA disputes some findings and is cooperating with the investigation.
tenfold has enhanced its free Community Edition Identity Governance tool with two new features: shared content governance and real-time event auditing. These additions help organizations with fewer than 150 users better manage Microsoft 365 sharing and monitor identity-related events. The shared content governance feature provides visibility into shared files across Teams, OneDrive, and SharePoint, along with recurring access reviews to control shared access. The event auditing feature enables real-time log analysis to detect suspicious identity activities promptly. This update aims to improve security and governance for small organizations without the complexity or cost of enterprise solutions.
SalesBleed was a zero-click attack chain disclosed by Zenity Labs that exploited Salesforce Agentforce's broad role-based permissions to leak CRM data via DNS exfiltration. The attack leveraged a poisoned lead submitted through a contact form, which the AI agent processed with excessive permissions, allowing data leakage despite patched parsing bugs. The root cause was an authorization design flaw where agents operate with broad role permissions without task-scoped authority, enabling shadow delegation attacks. This highlights the need for AI agents to have task-scoped permissions rather than broad role-based access to prevent misuse. Salesforce has fixed the specific bugs exploited, but the underlying authorization model risk remains in many agent deployments. The concept of a 'valet key'—scoped, limited authority per task—is proposed as a better security model for AI agents.
The US Senate has passed the bipartisan Health Care Cybersecurity and Resilience Act to strengthen cybersecurity in the healthcare sector. The Act aims to reduce successful cyberattacks, particularly ransomware and data extortion, which have severely impacted healthcare providers and patient data. It includes provisions for grants, training, improved interagency coordination, and updated regulations to enhance cybersecurity practices. The legislation now moves to the House of Representatives for consideration. Healthcare breaches have affected over 270 million Americans recently, with significant financial and operational impacts. The Act seeks to improve resilience and response capabilities across healthcare organizations, including rural providers.
OpenAI plans to introduce visual advertisements in ChatGPT during image generation sessions. This change represents an expansion of ads within the ChatGPT interface but does not describe a security vulnerability or threat.
The Microsoft Windows KB5124010 preview update released in September 2026 causes crashes in some games and applications that use AC-3 (Dolby Digital) audio decoding. The issue affects legacy applications relying on Windows built-in audio decoding, leading to failures to launch or unexpected closures when audio features are used. Modern applications that use their own audio decoding components are generally not impacted. The update is optional unless automatic updates are enabled on Windows 11 24H2 devices. Microsoft is aware of the problem and working on a fix but has not provided a timeline.
Google has suspended submissions to its Open Source Software Vulnerability Rewards Program (OSS VRP) due to a surge in AI-generated spam reports. This action is a response to the overwhelming volume of low-quality or automated vulnerability submissions, which impacts the program's effectiveness. No specific vulnerability or exploit is detailed in this report.
The alleged leader of the ShinyHunters extortion group, a teenage cybercriminal known as Rey (Saif al-Din Khader), was arrested in Jordan and is cooperating with the FBI to identify other members. The arrest followed a recent hack and defacement of the FBI's jobs site, with ShinyHunters claiming to have stolen terabytes of data including personal and health information of FBI employees. The FBI confirmed the group hacked over 140 organizations and extorted at least $70 million. Another suspect was arrested in the Netherlands as part of the investigation. The FBI continues to pursue additional leads and anticipates more arrests.
This entry is a daily update from the SANS Internet Storm Center (ISC) titled 'ISC Stormcast For Monday, October 5th, 2026.' It provides general information about the ISC's activities and links to a podcast but does not contain any specific security threat or vulnerability details.
This report describes an experiment involving the collection and analysis of TTY logs from actors or bots that successfully log into a DShield sensor. The logs capture commands executed by these actors, which are then parsed and sent daily to the DShield SIEM for correlation and further analysis. The data includes hashes of commands such as crontab entries executed by thousands of different IP addresses over a 90-day period. This activity provides insight into attacker behavior but does not describe a specific vulnerability or exploit.
The announcement comes after Trump hosted top executives of AI companies at the White House last week. The post Trump Names National Intelligence Director Jay Clayton to Lead a New Federal AI Task Force appeared first on SecurityWeek .
Post-quantum cryptography (PQC) standards were finalized by NIST in August 2024, replacing RSA and ECC for key exchange and digital signatures. While enterprise security teams are tracking PQC migration, most small and medium businesses (SMBs) and small SaaS companies are not, despite migration timelines spanning 3–10 years. This creates risk from adversaries who harvest encrypted data now to decrypt later once quantum capabilities mature. The urgency is especially high for organizations handling sensitive, long-lived data subject to regulations like HIPAA. A practical mitigation is to conduct a cryptographic inventory to identify asymmetric encryption usage, including in libraries and cloud defaults, and to consult vendor PQC roadmaps. Major cloud providers and OpenSSL have published PQC migration plans and experimental support. The migration at the TLS layer can often be done via configuration changes without code modifications.
Google's Gemini could soon access any file on your macOS device, open apps, browse the web, and perform actions without asking for permission every time. [...]
This content is a discussion post referencing a YouTube video that speculates about AI potentially reversing engineering digital systems and causing disruption. The post questions the video's claims, noting that actual compromise of server architectures would require encryption keys or authentication bypass methods. It also contrasts this with offline product piracy, which may be easier to achieve. There is no concrete vulnerability or exploit detailed.
YARA-X 1.21.0 is a software release that includes five improvements and four bug fixes. One notable enhancement is the ability to use standard input (stdin) for the command-line interface option --scan-list, enabling users to pipe a list of folders to scan. This release announcement does not describe any security vulnerabilities or threats.