Researchers say the new ‘Cryptographic Context Injection’ technique conceals malicious instructions until they are decrypted inside a trusted execution environment.

Grok 4.6 lands on Google Cloud Vertex AI with a 500K-token context window as Grok Bot expands to Windows and Linux with a ...

xAI's Grok web chat agent is currently vulnerable to a novel form of prompt injection, according to security researchers with ...

Adversa says encrypted prompt injection can make Grok send a user's name, location, tier, and chat prompts to an ...

The criminal AI service uses jailbreaks to bypass safeguards on legitimate AI models and offers capabilities including ...

Adversa AI says Grok still leaks user names, location, and full chat history to attackers hiding encrypted commands on a web ...

Elon Musk’s AI company xAI has put slightly more stringent limits on its Grok chatbot after it flooded the social app X with nonconsensual, highly sexualized AI-generated images depicting women, and ...

Elon Musk’s artificial intelligence software, Grok, continues to generate sexualized images of people without their consent, despite his company’s pledge months ago to halt abusive deepfakes after a ...

Grok, the built-in chatbot on X, is facing intense scrutiny after acknowledging it generated and shared an AI image depicting two young girls in sexualized attire. In a public post on X, Grok admitted ...