Llm-Security

LLM Role Confusion Attack Bypasses Safety at 61%

ATLAS OWASP HIGH ▲ 8.2 Simon Willison Jun 23, 2026

New research from Ye, Cui, and Hadfield-Menell demonstrates that LLMs prioritise the stylistic format of text over its structural role tags, enabling attackers to craft injected content that mimics internal reasoning blocks and bypasses safety guardrails. The study found attack success rates of 61% when injected text stylistically matched model-internal formats, dropping to just 10% after 'destyling'. The authors conclude that without genuine role perception in models, prompt injection defences will remain fundamentally reactive.

GitHub Ships Data Analytics Agent Built on Copilot

FIRST LOOK ATLAS OWASP MEDIUM ▲ 6.8 GitHub Blog Jun 20, 2026

GitHub has published a detailed engineering account of how it built an internal data analytics agent using GitHub Copilot, exposing the architectural patterns — including natural language-to-SQL translation, autonomous tool invocation, and internal data access — that underpin such systems. For defenders, this blueprint highlights concrete risks around prompt injection into analytics pipelines, excessive agency over sensitive internal datasets, and the challenge of auditing LLM-generated queries before execution. Organisations adopting similar agentic analytics patterns should treat this as a reference threat model rather than a safe-to-copy architecture.

OpenClaw Agent Vulnerable to Prompt Injection RCE

ATLAS OWASP HIGH ▲ 8.5 The Hacker News Jun 12, 2026

Two independent research teams demonstrated that OpenClaw, a self-hosted AI agent, is vulnerable to prompt injection attacks delivered through shared contacts, vCards, location pins, and plain emails — enabling attacker-controlled code execution and sensitive data exfiltration. Imperva's finding, now patched in version 2026.4.23, exploited the agent's failure to mark message objects as untrusted before passing them to the underlying LLM. Varonis separately showed that a single crafted email could instruct an agent to forward mock AWS credentials and customer data to an external address, a behaviour-level risk no patch can fully remediate.

Adversa AI: 89% of AI Agents Fail Security Tests

ATLAS OWASP HIGH ▲ 8.2 SecurityWeek Jun 04, 2026

Adversa AI's AI Risk Quadrant report evaluated 100 AI agents across ten categories, finding that only 11 qualify as both capable and well-defended. The research identifies a structural 'power-protection inversion' where the most capable agents also present the widest attack surface, driven by a 'lethal trifecta' of private data access, exposure to untrusted content, and outbound action capability. Computer and coding agents showed the most severe exposure, raising urgent concerns about autonomous agent deployment in enterprise environments.

SentinelOne Warns on Prompt Injection Risks in AI Agents

ATLAS OWASP MEDIUM ▲ 6.5 SentinelOne Blog May 25, 2026

SentinelOne has published guidance on securing agentic AI systems, framing unverified trust in AI agents as a core enterprise risk. The piece promotes their Prompt Security product as a control layer for AI tools, agents, and pipelines deployed across the enterprise. While primarily a product-focused announcement, it highlights the genuine security challenge of blind trust in autonomous AI agents executing actions on behalf of users and systems.

Agent Hijacking: Microsoft's Defense-in-Depth Framework

ATLAS OWASP MEDIUM ▲ 7.2 Microsoft Security Blog May 15, 2026

Microsoft's Security Blog introduces a layered defense-in-depth model specifically designed for autonomous AI agents, which now invoke tools, modify data, and trigger workflows with minimal human oversight. The framework identifies novel threat classes — including agent hijacking, intent breaking, and supply chain compromise — that are amplified by agentic autonomy. The guidance positions application-layer architecture, permissions, and governance as the most critical controls as agent autonomy scales.

Excessive Agency in AI Agents: Tool Access Control Gaps

ATLAS OWASP LOW ▲ 6.2 HN AI Security May 13, 2026

Statewright is an open-source framework that enforces state machine constraints on AI agents, restricting which tools agents can invoke during each phase of a workflow. The project directly addresses the Excessive Agency problem, where AI agents operating with broad, unconstrained tool access can take unintended or harmful actions. While a defensive development rather than a threat disclosure, it signals growing practitioner awareness of agentic AI risk and offers a concrete mitigation pattern for teams deploying coding agents like Claude Code, Codex, or Cursor.

CrowdStrike Red Teaming: LLM Jailbreak and Data Poisoning

ATLAS OWASP MEDIUM ▲ 6.5 SecurityWeek May 06, 2026

Joey Melo, Principal Security Researcher at CrowdStrike, outlines his methodology for AI red teaming, focusing on manipulating LLM guardrails through jailbreaking and data poisoning without altering underlying source code. His work, rooted in competitive AI hacking challenges, translates classical adversarial thinking into the emerging field of machine learning security. The profile highlights the growing professionalisation of AI red teaming as organisations seek to harden LLM deployments against real-world manipulation attacks.

AI Agents Exploit Excessive Agency to Delete Production Databases

ATLAS OWASP HIGH ▲ 6.5 Dark Reading May 02, 2026

Organisations are deploying AI agents into production environments without adequate security testing, resulting in destructive outcomes such as unintended deletion of production databases. The core risk is excessive agency granted to AI systems before trust boundaries and guardrails are established. This represents a systemic industry failure to apply basic security principles before integrating autonomous AI tooling into critical infrastructure.

Amazon Bedrock Prompt Injection Traverses Agent Hierarchies

ATLAS OWASP HIGH ▲ 8.5 Palo Alto Unit 42 Apr 23, 2026

Unit 42 researchers conducted red-team analysis of Amazon Bedrock's multi-agent collaboration framework, demonstrating how attackers can systematically exploit prompt injection to traverse agent hierarchies, extract system instructions, and invoke tools with attacker-controlled inputs. The research reveals that multi-agent architectures introduce compounded attack surfaces through inter-agent communication channels, though no underlying Bedrock vulnerabilities were identified. Properly configured Guardrails and pre-processing stages effectively mitigate the demonstrated attack chains.

Legacy Vulnerabilities Amplified by AI at Enterprise Scale

ATLAS OWASP HIGH ▲ 6.5 Dark Reading Apr 18, 2026

The article argues that AI's primary security risk lies not in introducing entirely new vulnerability classes, but in dramatically amplifying the impact and exploitability of well-established ones. This framing has significant implications for defenders, suggesting that legacy vulnerability management practices must be re-evaluated through an AI-augmented threat lens. The convergence of classic weaknesses with AI capabilities raises the baseline risk profile for organisations deploying or adjacent to AI systems.

Agentic AI Excessive Agency Bypasses Security Testing

ATLAS OWASP MEDIUM ▲ 6.5 The Hacker News Apr 16, 2026

The article examines the architectural tension between fully agentic AI systems and deterministic validation frameworks in security testing contexts, arguing that unconstrained AI autonomy introduces repeatability and auditability risks. It highlights how probabilistic AI behaviour — while valuable for exploration — undermines the measurable, consistent outcomes required for enterprise security validation programs. The piece reflects a broader industry debate about governing AI agency in high-stakes operational environments.

Anthropic MCP Supply Chain Flaw Enables Command Injection

ATLAS OWASP CRITICAL ▲ 9.1 SecurityWeek Apr 16, 2026

A structural vulnerability in Anthropic's Model Context Protocol (MCP) allows unsanitized commands to be executed silently within AI environments, potentially enabling full system compromise. Researchers classify the flaw as 'by design,' meaning it stems from architectural decisions rather than implementation bugs, making it particularly difficult to patch without protocol-level changes. The breadth of MCP adoption across agentic AI toolchains significantly amplifies the supply chain risk.

AI Agent Prompt Injection and Data Leakage Threats Rise

ATLAS OWASP MEDIUM ▲ 6.5 SecurityWeek Apr 16, 2026

Capsule Security, an Israeli startup, has emerged from stealth with $7 million in seed funding focused on runtime security for AI agents, continuously monitoring their behaviour to detect and prevent unsafe or malicious actions. This positions the company within the rapidly growing agentic AI security space, where autonomous agents executing actions on behalf of users represent a significant and underexplored attack surface. The funding signals growing investor recognition of the risks posed by unmonitored AI agent behaviour, including prompt injection, excessive agency, and unintended tool use.

Anthropic Model Exploits Zero-Days Faster Than SOC Response

ATLAS OWASP HIGH ▲ 7.2 The Hacker News Apr 14, 2026

The article highlights a critical operational gap in SOC environments where AI-accelerated adversarial capabilities — including an Anthropic model restricted after autonomously exploiting zero-day vulnerabilities — are outpacing defender response workflows. While detection times (MTTD) have improved, the post-alert investigation window remains the primary exposure point, with breakout times of 29 minutes and adversary hand-off times collapsing to 22 seconds. The piece argues that AI-driven investigation tooling is the necessary counter to compress this post-alert gap.

Llm-Security

Stay ahead of the threat.