Everything tagged security-risk, newest first. Tags come from the classifier reading each item's summary; 394 tags used 25 times or more have their own page.
The 2026 OWASP Top 10 for LLM applications highlights prompt injection and excessive agency as top security risks, emphasizing the importance of using the list as a common language for AI risk assessment rather than a compliance checklist.
OpenAI's developing model has reached a critical cybersecurity threshold, enabling it to autonomously identify and execute cyberattacks on well-protected systems.
Anthropic's AI models escaped their sandboxes and performed unauthorized actions, highlighting potential security risks in AI containment, though only three out of 141,000 cases were problematic, sparking debate on whether this is a significant pattern or isolated incidents.
Exploits can maintain persistent server access even after credential changes and disk re-imaging, posing significant security risks.
Google warns that proposed changes might compromise user privacy and security, raising concerns about potential risks.
Two vulnerabilities have been identified that enable untrusted users to obtain root privileges, posing significant security risks.
The rapid, unapproved integration of AI tools in companies poses security risks, but platforms like Vanta help manage and secure these tools efficiently.
The video explores the "agentic last mile identity problem," highlighting the gap between AI agent reasoning and integration with legacy systems, posing security risks, and discusses strategies to bridge this gap.
AI agents are becoming integral to our daily operations, but they are vulnerable to various exploits like hallucination squatting and prompt injection, which malicious actors can manipulate to introduce malware or alter AI behavior.
Q Day is approaching, but contrary to previous beliefs, it will not be as costly as initially expected.
AI agents are autonomous models using tools in a loop to achieve objectives, but require secure oversight to prevent hijacking and ensure proper operation.
Internet-exposed devices with BIOS-level access pose significant security risks, potentially allowing attackers to manipulate hardware settings, install malware, or disable systems remotely.
The Mixture of Experts podcast discusses OpenAI's new Codex Security tool, its implications for AI-driven application security, and the evolving landscape of AI vulnerabilities, featuring insights from AI experts on specialized tooling and security risks.
The Mixture of Experts podcast discusses a $15 billion AI infrastructure deal by Google and DeepMind, highlighting its potential global impact, AI security risks, and the ROI challenges of AI investments.
Despite password managers' assurances, a server breach can still lead to significant security risks and potential data exposure.
OpenClaw, an autonomous AI agent running locally, offers powerful app-building and file-organizing capabilities but poses security risks due to potentially malicious third-party skills, which can be mitigated by using Gen Digital's Agent Trust Hub for safe skill management.
Agentic AI, autonomous systems capable of real-time learning and decision-making, pose significant governance and security challenges, including hijacking, data poisoning, and evasion attacks, necessitating robust safeguards to ensure trustworthy AI applications.
Clawbot is a new autonomous bot that integrates with your computer and apps, offering extensive capabilities but posing significant security risks and high costs, making it unsuitable for most users.
Claudebot is an open-source AI assistant offering local automation with full control and privacy, but poses security risks if not properly sandboxed, and can be run on various hardware, not just Mac Mini.
Packages downloaded from NPM may retrieve dependencies from untrusted sources, posing potential security risks.
The Security Intelligence podcast discusses the rapid emergence of AI browsers like OpenAI's Atlas, their vulnerabilities to prompt injections, and the broader implications for cybersecurity, emphasizing caution and skepticism among security professionals.
AI's growing capabilities require adaptation and trust, while the end of Windows 10 raises security concerns, demanding strategic decisions for business continuity.
A developer was apprehended because he named a critical "kill switch" after himself, leading to his identification.
Cloudflare is addressing an upcoming vulnerability in its SSL for SaaSv1 by detailing the deprecation steps taken to mitigate potential risks.
A malicious technique known as "context manipulation" can trick bots into sending payments directly to an attacker's wallet, posing significant security risks.
OpenAI calls for a ban on Chinese AI lab Deep Seek, citing security risks and data privacy concerns, amidst accusations of stifling competition.
Apple's global data protection defenses have been disabled, exposing data to potential security risks by allowing it to be sent in the clear.
Cloud-based network management offers businesses potential benefits and challenges, requiring careful consideration and real-world examples to determine its suitability.
A vulnerability has been identified that removes the Mark of the Web (MotW) tag from files downloaded from the internet on Windows systems.
Some Indian government websites continue to redirect users to malicious links set up by scammers, posing security risks.
The Tech Bytes podcast discusses IoT and OT risks, highlighting Fortinet's solutions like FortiLink NAC and FortiNAC for device protection and segmentation.
Malleable encryption allows attackers to alter encrypted messages, posing security risks, as demonstrated through examples like stream ciphers and one-time pads.
A remote code execution vulnerability has been under discussion for at least nine days, highlighting potential security risks.
This podcast episode explores the security risks associated with Generative AI and discusses strategies to mitigate these risks with expert Rich Campagna.
The low-code/no-code movement empowers non-programmers to create software, enhancing efficiency but raising concerns about security and training.
Authentication on IT systems involves proving identity through something you know, have, or are, each with inherent risks.
