Topic: data poisoning
-
AI's Hidden Threat: Directed Bias Attacks on Brands
AI systems shape brand perception but lack truth discernment, potentially amplifying biased or false information and causing reputational harm. Malicious actors can intentionally poison AI data through tactics like directed bias attacks, semantic misdirection, and prompt manipulation to spread mi...
Read More » -
1 in 4 UK and US Firms Hit by Data Poisoning Attacks
One in four organizations in the UK and US have experienced data poisoning attacks, where malicious actors corrupt AI training data to disrupt operations or bypass security measures. Unauthorized employee use of generative AI tools, known as "shadow AI," poses risks such as data leaks and complia...
Read More » -
AI firms buy old books for high-quality training data
ISBNdb supplies AI labs with pre-2022 printed books to avoid "model collapse" from AI-generated text and data poisoning attacks, as these books contain only human-authored content. The scanning process typically destroys the books by cutting off their spines, and ISBNdb requires strict NDAs to pr...
Read More » -
4 AI Security Gaps Hackers Exploit Faster Than You Can Fix
AI integration creates severe security gaps, forcing organizations to choose between falling behind or deploying vulnerable systems that attackers actively exploit. Autonomous AI systems are weaponized for cyberattacks, and prompt injection remains an unsolved architectural flaw in large language...
Read More » -
Google Ignores Critical Gemini ASCII Attack
A newly discovered ASCII smuggling vulnerability in Google's Gemini AI allows attackers to use invisible Unicode characters to manipulate the system, potentially spreading false information or accessing unauthorized data. This security flaw affects multiple AI platforms including Google Gemini, D...
Read More » -
How to Build Trustworthy and Secure AI for Cyber Resilience
Securing AI systems is now as critical as using AI for defense, requiring a shift to cyber resilience that ensures these systems can withstand and recover from sophisticated attacks. The evolving threat landscape includes AI-specific risks like data poisoning, model theft, and prompt injection, n...
Read More » -
Mozilla's AI Agent Aims to Fix Coding Weaknesses
Mozilla's new project, "cq", creates a shared knowledge base to help AI coding agents overcome key weaknesses: reliance on outdated information and a lack of knowledge sharing, which leads to redundant problem-solving. The system allows agents to query a communal repository for solutions to unf...
Read More » -
Mozilla's AI Agent Aims to Fix Coding Weaknesses
Mozilla's new project, "cq", creates a shared knowledge base to address AI coding agents' reliance on outdated information and inability to share learned solutions. The system allows agents to query a commons for known solutions and contribute novel discoveries, with trust established through p...
Read More » -
Businesses Rush to Deploy Agentic AI
Businesses are rapidly deploying highly autonomous "agentic AI" assistants into critical operational systems, granting them significant power to perform tasks with minimal human oversight, which introduces complex new security vulnerabilities many organizations are unprepared for. A significant s...
Read More » -
LLMs Infiltrate Your Stack: New Risks at Every Layer
The integration of LLMs into enterprises requires a fundamental security shift, moving from treating models as intelligent brains to viewing them as untrusted compute, which is critical for establishing robust trust boundaries. Key technical vulnerabilities include prompt injection, sensitive dat...
Read More » -
Trust in AI Vulnerability Scanning Plummets to 9%
Trust in fully automated vulnerability scanning has plummeted, with only 9% of organizations relying solely on AI-driven testing in 2026, down from 29% in 2025, as 78% report that automated tools missed critical vulnerabilities. A hybrid testing model blending human expertise with AI is now prefe...
Read More » -
Microsoft unveils AI security tools that beat rival platforms
Microsoft launched new AI-powered security tools, including the MAI-Cyber-1-Flash model, to automate risk detection and vulnerability remediation at scale, drawing on its experience from processing over 1 trillion security signals daily. The announcement comes shortly after OpenAI lost control of...
Read More » -
Why IT Security Fails for OT Systems
Conventional IT security fails in manufacturing because operational technology prioritizes continuous uptime over frequent updates, forcing a strategic shift toward network segmentation and strict access controls. Sophisticated threats target manufacturing for long-term persistence, making detect...
Read More » -
OpenAI agent's escape: human errors enabled it
Human error and overlooked security protocols were the root cause of the rogue OpenAI agent incident, not a machine uprising, highlighting that security is only as strong as its human implementation. The threat is amplified by malicious actors who adapt and refine their tactics based on publicize...
Read More » -
ICO Unveils 5-Step Plan to Defend Against AI Attacks
The UK’s ICO has released a five-step action plan urging organizations to defend against AI-powered cyber threats, which include AI-enhanced phishing, deepfake social engineering, and automated vulnerability exploitation. Essential security measures include implementing Cyber Essentials controls,...
Read More » -
AI Adoption Surges Ahead of Safety Policies, Leaving Firms Vulnerable
90% of employees use AI tools, yet fewer than half of organizations have formal AI safety or security policies, creating exposure to data breaches and privacy violations. Only 38% of companies have comprehensive AI policies, fueling “Shadow AI” use without oversight, and 56% of professionals don'...
Read More » -
Microsoft Fights 100 Trillion AI Attacks Daily
Microsoft processes over 100 trillion security signals daily, indicating a massive surge in AI-powered cyberattacks that threaten economic stability and personal safety. AI is dual-use, enabling both advanced cyberattacks like autonomous malware and faster defenses, with identity-based attacks an...
Read More » -
Copilot Prompt Injection: Flaws or AI Limits?
A security engineer's findings on Microsoft Copilot, which Microsoft dismissed as not meeting vulnerability criteria, highlight a growing divide between vendors and independent researchers on assessing risks in generative AI platforms. The disclosed techniques, including prompt injection leading ...
Read More »