Loading story…
OWASP’s GenAI Security Project published the 2026 Top 10 for LLM Applications, updating its guidance for securing AI applications and autonomous agents with a methodology that blends community input and real-world incident data. The project analyzed 7,714 AI-related security incidents, including 6,639 detailed enough for classification, making this the first edition shaped in part by observed incidents rather than practitioner consensus alone. Prompt Injection remained the top-ranked risk, Sensitive Information Disclosure stayed near the top, and Excessive Agency rose to third as agentic deployments showed growing operational risk.
The update also reshaped several categories to reflect how attacks and failures are evolving. OWASP renamed System Prompt Leakage to Hidden Context Exposure, expanded Prompt Injection to include cross-modal attacks, and broadened Data and Model Poisoning to cover fine-tuning subversion. Misinformation moved higher because incident data showed significant real-world harm, while Unbounded Consumption was highlighted as an emerging availability and cost risk. OWASP said organizations should build resilient architectures around fallible models, using containment, least agency, access control before retrieval, output validation, and supply-chain security, and mapped the risks to frameworks including MITRE ATLAS, MITRE ATT&CK, CWE, NIST AI 600-1, NIST AI RMF, and the CSA AI Controls Matrix.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
4 events from the most recent confirmed update back to the earliest known activity.
OWASP rolled out a new agentic AI security maturity framework in June 2026 to help organizations align deployed agentic systems with required governance. The framework was referenced alongside the 2026 LLM Top 10 update.
OWASP renamed System Prompt Leakage to Hidden Context Exposure and broadened it to cover non-user-visible contexts such as system instructions, RAG schemas, and hidden policy logic. It also expanded Prompt Injection to include cross-modal attacks and broadened Data and Model Poisoning to include fine-tuning subversion.
The 2026 edition became the first OWASP LLM Top 10 influenced by real-world incident data rather than practitioner consensus alone. OWASP weighted practitioner voting at about 75% and incident data at 25%, using 6,639 classifiable incidents from a larger dataset of 7,714 AI-related incidents.
OWASP's GenAI Security Project released the 2026 edition of its Top 10 for LLM Applications. The update kept Prompt Injection as the top risk and Sensitive Information Disclosure second while revising lower-ranked categories based on community input and incident data.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
7 references tracked. Mallory keeps watching after this page renders.
heise.de
Open sourcereversinglabs.com
Open sourcescworld.com
Open sourcecybersecuritynews.com
Open sourcehelpnetsecurity.com
Open sourceinfosecurity-magazine.com
Open sourcegenai.owasp.org
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
The Open Worldwide Application Security Project (OWASP) has launched a comprehensive initiative to address the security risks associated with agentic and autonomous AI systems. This includes the release of the "Top 10 for Agentic Applications 2026," a globally peer-reviewed framework that identifies the most critical security risks for these AI systems, along with an AI testing guide and a dynamic web-based vulnerability assessment tool. The framework highlights risks such as agent goal hijack and tool misuse, providing organizations with actionable mitigation recommendations and practical tools to secure their AI deployments. These efforts come as industries, particularly financial services, face increasing threats from generative and agentic AI, which can accelerate attack timelines and introduce new vectors such as deepfake-driven fraud and rapid ransomware campaigns. The OWASP initiative aims to equip security teams with the necessary resources to proactively address these evolving risks, emphasizing the importance of embedding security into AI systems from the outset and keeping pace with the rapidly changing threat landscape.
Jun 29, 2026
The Open Web Application Security Project (OWASP) has officially published its inaugural list of the top ten security threats facing agentic artificial intelligence (AI) applications. Announced at Black Hat Europe 2025, the list highlights risks such as agent goal hijacking, privilege abuse, unexpected code execution, insecure inter-agent communication, and memory/context poisoning. Alongside the list, OWASP released governance and security guides for AI agents, a visual risk map for open-source and commercial agentic AI tools, and a Capture The Flag application (FinBot) to help cybersecurity teams practice defending against these threats. The initiative aims to help organizations understand and mitigate the rapidly expanding attack surface introduced by the proliferation of AI agents across enterprise environments. Industry experts, including members of the Agentic Security Initiative (ASI) Distinguished Review Board, have emphasized the significance of this release, noting the growing adoption of agentic AI in sectors such as GRC, AppSec, and SecOps, as well as its exploitation by malicious actors. The OWASP Top 10 for Agentic Applications is positioned as a foundational resource for security leaders, developers, and practitioners to assess and address the unique risks posed by autonomous AI agents, supplementing previous OWASP projects focused on web applications and large language models. The publication is expected to drive further research, awareness, and best practices in securing agentic AI systems as their use becomes more widespread.
Jun 29, 2026
The Open Web Application Security Project (OWASP) has published its inaugural "Top 10 for Agentic Applications," a risk framework specifically addressing the unique security challenges posed by advanced AI agents. These agents, which go beyond simple chatbots to autonomously access data, use tools, and execute tasks, introduce new attack surfaces and risks such as agent goal hijacking, tool misuse, and privilege abuse. The list was developed with input from over 100 security researchers and validated by experts from organizations like NIST and the European Commission, and is based on real-world incidents where AI agents have been manipulated to exfiltrate data, misuse tools, or cause cascading failures in enterprise workflows. Security experts highlight that the rise of agentic AI also exposes previously overlooked vulnerabilities, such as shadow APIs and legacy systems that were once considered secure by obscurity. AI agents are capable of discovering and interacting with undocumented or forgotten APIs, making even antiquated systems vulnerable if they are connected to modern networks. The new OWASP framework underscores the urgent need for organizations to reassess their security postures, increase visibility into internal systems, and proactively address the risks introduced by autonomous AI agents and their ability to exploit both new and legacy infrastructure.
Jun 29, 2026