#prompt-injection
#prompt-injection

Privacy professionals

Hackers exploit a blind spot by hiding malware inside DNS records

fromHackernoon

3 weeks ago

Privacy professionals

The Prompt Protocol: Why Tomorrow's Security Nightmares Will Be Whispered, Not Coded | HackerNoon

Privacy professionals

Hackers exploit a blind spot by hiding malware inside DNS records

more#ai-vulnerabilities

#peer-review

Artificial intelligence

Scientists Are Sneaking Passages Into Research Papers Designed to Trick AI Reviewers

fromNature

Privacy professionals

Scientists hide messages in papers to game AI peer review

Artificial intelligence

Scientists Are Sneaking Passages Into Research Papers Designed to Trick AI Reviewers

fromNature

Privacy professionals

Scientists hide messages in papers to game AI peer review

Artificial intelligence

Google Adds Multi-Layered Defenses to Secure GenAI from Prompt Injection Attacks

fromTheregister

2 weeks ago

Artificial intelligence

Scholars sneaking phrases into papers to fool AI reviewers

1 month ago

Artificial intelligence

Google Adds Multi-Layered Defenses to Secure GenAI from Prompt Injection Attacks

fromTheregister

2 weeks ago

Artificial intelligence

Scholars sneaking phrases into papers to fool AI reviewers

more#machine-learning

#ai

Artificial intelligence

GitLab Duo Vulnerability Enabled Attackers to Hijack AI Responses with Hidden Prompts

Artificial intelligence

Researchers claim breakthrough in fight against AI's frustrating security hole

fromTechzine Global

1 month ago

Artificial intelligence

Zero-click attack reveals new AI vulnerability

Artificial intelligence

Researchers cause GitLab AI developer assistant to turn safe code malicious

GitLab Duo Vulnerability Enabled Attackers to Hijack AI Responses with Hidden Prompts

GitLab's AI assistant Duo has a serious security vulnerability that could expose sensitive information and enable code theft.

Researchers claim breakthrough in fight against AI's frustrating security hole

Prompt injections jeopardize AI systems; Google DeepMind's CaMeL offers a potential solution by treating language models as untrusted components within security frameworks.

fromTechzine Global

1 month ago

Zero-click attack reveals new AI vulnerability

Echoleak exposes vulnerabilities in AI assistants like Microsoft 365 Copilot through subtle prompt manipulation, representing a shift in cybersecurity attack vectors.

Artificial intelligence

Researchers cause GitLab AI developer assistant to turn safe code malicious

more#ai

#ai-security

5 months ago

Google Gemini's Long-term Memory Vulnerable to a Kind of Phishing Attack

Johann Rehberger demonstrated a prompt injection attack on Google Gemini, exploiting delayed tool invocation to modify its long-term memories.

Researchers Find Easy Way to Jailbreak Every Major AI, From ChatGPT to Claude

A newly discovered jailbreak can manipulate AI models into producing harmful content, exposing vulnerabilities in their safety measures.

DeepMind Researchers Propose Defense Against LLM Prompt Injection

Google DeepMind's CaMeL effectively neutralizes 67% of prompt injection attacks in LLMs using traditional software security principles.

Growth hacking

Gemini hackers can deliver more potent attacks with a helping hand from... Gemini

Indirect prompt injections are an effective method for exploiting large language models, revealing vulnerabilities in AI systems.

Artificial intelligence

Meta Open Sources LlamaFirewall for AI Agent Combined Protection

Artificial intelligence

Researchers Demonstrate How MCP Prompt Injection Can Be Used for Both Attack and Defense

5 months ago

Google Gemini's Long-term Memory Vulnerable to a Kind of Phishing Attack

Johann Rehberger demonstrated a prompt injection attack on Google Gemini, exploiting delayed tool invocation to modify its long-term memories.

Researchers Find Easy Way to Jailbreak Every Major AI, From ChatGPT to Claude

A newly discovered jailbreak can manipulate AI models into producing harmful content, exposing vulnerabilities in their safety measures.

DeepMind Researchers Propose Defense Against LLM Prompt Injection

Google DeepMind's CaMeL effectively neutralizes 67% of prompt injection attacks in LLMs using traditional software security principles.

Growth hacking

Gemini hackers can deliver more potent attacks with a helping hand from... Gemini

Indirect prompt injections are an effective method for exploiting large language models, revealing vulnerabilities in AI systems.

Artificial intelligence

Meta Open Sources LlamaFirewall for AI Agent Combined Protection

Artificial intelligence

Researchers Demonstrate How MCP Prompt Injection Can Be Used for Both Attack and Defense

more#ai-security