Zenity Labs
PulseAugur coverage of Zenity Labs — every cluster mentioning Zenity Labs across labs, papers, and developer communities, ranked by signal.
2 day(s) with sentiment data
-
Salesforce Agentforce Vulnerable to Prompt Injection via Public Forms
A security vulnerability dubbed SalesBleed has been disclosed, allowing malicious prompts to be injected through public-facing forms and exfiltrate sensitive data from CRM systems. Researchers at Zenity Labs discovered …
-
Prompt Injection Vulnerability Found in AWS Bedrock AgentCore
Zenity Labs has demonstrated a critical vulnerability in AWS Bedrock AgentCore, where a single prompt injection could allow an attacker to gain broad access to an account's resources. This exploit could enable an attack…
-
Microsoft integrates local AI models into Windows ML; AWS agents vulnerable to prompt injection
Microsoft has integrated support for llama.cpp into Windows ML, allowing developers to run GGUF format AI models directly on Windows 11 without conversion to ONNX. This experimental feature simplifies local AI developme…
-
AgentCorruption vulnerability collapses cloud security AI models
A security vulnerability dubbed "AgentCorruption" has been discovered, capable of collapsing entire cloud security models with a single prompt. This exploit, detailed by Zenity Labs, highlights a critical flaw in how AI…
-
Zenity Labs exploits Amazon AgentCore with single prompt
Zenity Labs has demonstrated a security vulnerability within Amazon's AgentCore system, where a single prompt could compromise internal AWS interfaces. This exploit highlights a lack of strict isolation, potentially all…
-
Zenity researchers find AWS AI agents vulnerable to account-wide takeover
Researchers at Zenity Labs have discovered a critical vulnerability in Amazon Web Services' Bedrock AgentCore. A single, publicly accessible AI agent was found to be capable of compromising all other AgentCore agents wi…
-
AI agents struggle with rule discovery; test-time training trade-offs explored
A new benchmark, dig.bench, has been released to evaluate AI agents' ability to discover unknown game rules through experimentation, with current top models struggling to match human performance. Separately, research in…
-
AI agents can become persistent insider threats, new research shows
A new attack method called AgentForger has been developed by Zenity Labs, demonstrating that AI agents can pose persistent insider threats. This method highlights the potential for AI agents to be exploited for maliciou…
-
AgentForger vulnerability allows malicious AI agents via manipulated ChatGPT links
A security vulnerability named AgentForger has been discovered, allowing attackers to remotely install malicious AI agents within corporate networks. This exploit leverages manipulated links, specifically targeting Chat…