#ai-assistant-security

[ follow ]
#openclaw #sandbox-escape #privilege-escalation #cve-vulnerabilities #virustotal #prompt-injection
Information security
fromSecurityWeek
16 hours ago

'Claw Chain' OpenClaw Flaws Allow Sandbox Escape, Backdoor Delivery

Chained OpenClaw vulnerabilities let attackers with sandbox code execution control the agent, bypass restrictions, steal secrets, escalate privileges, and persist on the host.
Information security
fromTheregister
3 months ago

Senator doesn't trust telcos on Salt Typhoon mitigations

OpenClaw now scans ClawHub skills with VirusTotal across 70+ engines for malware and blocklists, but signature-based scanning cannot stop prompt injection or language-driven attacks; telco breach scrutiny continues.
Information security
fromInfoQ
5 months ago

Securing AI Assistants: Strategies and Practices for Protecting Data

Protecting data across AI assistant pipelines—from ingestion, transformation, to deployment and monitoring—is critical to secure operations.
Information security
fromIT Pro
7 months ago

A malicious MCP server is silently stealing user emails

A malicious MCP server repackaged as Postmark on npm exfiltrated thousands of emails by adding a BCC line, exploiting full assistant privileges and bypassing security controls.
[ Load more ]