Moltbook Network Exposes Bot-to-Bot Data Leaks

A new investigation into the Moltbook network exposes data vulnerabilities and raises alarm over prompt injection risks between AI agents. Researchers from Wiz and Permiso analyzed the AI-driven social platform and discovered multiple security flaws. These include bot-to-bot prompt injection attacks, where one agent can manipulate another’s behavior using crafted text prompts, and unintentional data leaks that could compromise sensitive information.

Researchers noted that Moltbook’s architecture allows interconnected bots to share and respond to outputs — a design choice that opens pathways for malicious actors. Attackers could embed malicious prompts in shared content streams, triggering unintended behaviors across the network.

The findings highlight the potential consequences of insufficient guardrails in agent-based AI systems. As such platforms gain traction, security analysts emphasize proactive auditing and containment mechanisms. The report concludes that the Moltbook network exposes data in ways that put broader AI ecosystems at risk.

Read the full analysis here: https://www.securityweek.com/security-analysis-of-moltbook-agent-network-bot-to-bot-prompt-injection-and-data-leaks/