Anthropic published the prompt injection failure rates that enterprise security teams have been asking every vendor for

Anthropic's Opus 4.6 system card breaks out prompt injection attack success rates by surface, attempt count, and safeguard ...

These 4 critical AI vulnerabilities are being exploited faster than defenders can respond

These 4 critical AI vulnerabilities are being exploited faster than defenders can respond ...
Morning Overview on MSN

Anthropic warns AI risk of committing heinous crimes is 'not negligible'

Anthropic has acknowledged in a formal transparency disclosure that the probability of advanced AI systems enabling or committing serious crimes is “not negligible,” a phrase that carries significant ...
VentureBeat

Why GPT-4 is vulnerable to multimodal prompt injection image attacks

OpenAI's new GPT-4V release supports image uploads — creating a whole new attack vector making large language models (LLMs) vulnerable to multimodal injection image attacks. Attackers can embed ...

Viral AI Caricatures Highlight Shadow AI Dangers

A viral AI caricature trend may be exposing sensitive enterprise data, fueling shadow AI risks, social engineering attacks, ...
SlashGear

AI Browsers Face A New Kind Of Attack, And It Puts Your Privacy At Risk

Companies like OpenAI, Perplexity, and The Browser Company are in a race to build AI browsers that can do more than just display webpages. It feels similar to the first browser wars that gave us ...

LLM Security Isn’t Just Theoretical—It’s A QA Problem You Can Test

As a QA leader, there are many practical items that can be checked, and each has a success test. The following list outlines what you need to know: • Source Hygiene: Content needs to come from trusted ...
India Today on MSN

ChatGPT will now tell you when your private data may be at most risk of leaking

OpenAI has rolled out Lockdown Mode and new Elevated Risk labels in ChatGPT to help users keep their sensitive information safer. According to the company, these tools will warn users about ...
Futurism

OpenAI’s New AI Browser Is Already Falling Victim to Prompt Injection Attacks

OpenAI unveiled its Atlas AI browser this week, and it’s already catching heat. Cybersecurity researchers are particularly alarmed by its integrated “agent mode,” currently limited to paying ...
MIT Technology Review

Is a secure AI assistant possible?

AI agents are a risky business. Even when stuck inside the chatbox window, LLMs will make mistakes and behave badly. Once ...
Security Boulevard

OpenClaw Open Source AI Agent Application Attack Surface and Security Risk System Analysis

Background In early 2026, OpenClaw (formerly known as Clawdbot and Moltbot), an open-source autonomous AI agent project, quickly attracted global attention. As an automated intelligent application ...
InfoWorld

Single prompt breaks AI safety in 15 major language models

The GRP‑Obliteration technique reveals that even mild prompts can reshape internal safety mechanisms, raising oversight concerns as enterprises increasingly fine‑tune open‑weight models with ...