CrewAI has integrated native memory protection into its agent configuration, addressing critical security vulnerabilities related to persistent agent memory. This feature aims to prevent malicious injections that could lead to privilege escalation or data exfiltration, directly tackling OWASP Top 10 risks.
This content identifies two DM-hardening problems in AI systems, focusing on the security aspect where hostile DMs exploit LLM reasoning to trigger unauthorized actions. It presents `v0.21`'s solution: a provenance layer that validates actions based on DM origin rather than content.
Traditional web application security taxonomies like MITRE ATT&CK do not cover new threats specific to Large Language Models (LLMs), such as prompt injection or jailbreaks. MITRE ATLAS was developed to bridge this gap, providing a standardized framework for threats in AI/ML systems.
Prompt injection is a security vulnerability where untrusted input is interpreted as instructions by an LLM, allowing attackers to override system behavior. Effective prevention requires pre-LLM sanitization of user input through validation and filtering, ideally using code-level static analysis rather than just runtime filters.
This article explores engineering lessons from a hypothetical 16-million-conversation leak involving Anthropic's Claude, emphasizing LLMs as a distinct attack surface. It highlights that threat boundaries now encompass third-party environments and rich logs, expanding a chat breach into a business breach.
As AI systems rapidly integrate into enterprise operations, security becomes a critical concern. AI red team testing is essential to identify vulnerabilities and new attack surfaces that traditional testing methods fail to address in dynamic models.
This article explores the risk of indirect prompt injection in AI agents, such as Claude Desktop, that use external tools via the Model Context Protocol (MCP). It details how tool outputs can become attack vectors and offers code solutions to implement injection scanning.
This content explains how to secure the Claude CLI when it runs server-side within your software, addressing prompt injection risks. It details the use of security layers like `--print` and `--bare` to disable interactive tools and reduce available context to the model.
TEOS Sentinel Shield is an AI execution firewall that blocks unsafe AI code, including LLM-generated scripts, before it runs. It analyzes 14+ risk patterns in under 2 seconds, providing clear ALLOW/WARN/BLOCK decisions to prevent system compromises.
The content announces upcoming workshops at NDC Sydney 2026 and Black Hat Asia, focusing on AI security, exploring LLM vulnerabilities and secure design patterns. It includes an intense AI attack and defence wargame where participants secure and attack AI chatbots.