[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"framework:mitre-atlas\u002Faml-t0054":3},{"item":4,"lessons":16,"controls":54,"related":82},{"family":5,"item":6,"code":7,"title":8,"citations":9,"publisher":10,"summary_md":11,"official_text":12,"official_attribution":12,"official_url":13,"link_label":14,"note":12,"updated_at":15},"mitre-atlas","aml-t0054","AML.T0054","LLM Jailbreak",9,"MITRE","**What attackers do.** This item maps to MITRE ATLAS, the knowledge base for attacks on AI systems, where the technique covers crafted prompts meant to push a large language model outside its intended behaviour and safety controls. Attackers use role play, encoding tricks or layered instructions so the model produces content or takes actions it should refuse.\n\n**How to defend.** Do not rely on model guardrails alone: add independent input and output filtering, restrict what tools and data the model can use, and apply least privilege to any actions it can take. Keep logs of prompts and responses, red team applications with known bypass patterns, and update protections as new methods appear.\n\n**Signs to look for.** Look for prompts with unusual encodings, long role play setups, repeated rephrasing after refusals, and outputs that violate policy. A common gap is monitoring only for harmful output while ignoring repeated probing attempts.",null,"https:\u002F\u002Fatlas.mitre.org\u002F","Open MITRE ATLAS","2026-10-18T15:18:43.663+00:00",{"total":9,"list":17},[18,22,26,30,34,38,42,46,50],{"slug":19,"title":20,"published_at":21},"single-ai-prompt-could-compromise-entire-aws-environment-via-agentcore-flaw","Single AI Prompt Could Compromise Entire AWS Environment via AgentCore Flaw","2026-10-08T22:21:21.853+00:00",{"slug":23,"title":24,"published_at":25},"ai-chatbots-weaponized-via-web-content-poisoning-for-phishing-disinformation","AI Chatbots Weaponized via Web Content Poisoning for Phishing & Disinformation","2026-09-23T21:21:28.203+00:00",{"slug":27,"title":28,"published_at":29},"nvidia-nemoclaw-flaw-lets-malicious-webpages-hijack-local-ai-models","NVIDIA NemoClaw Flaw Lets Malicious Webpages Hijack Local AI Models","2026-08-25T16:21:20.787+00:00",{"slug":31,"title":32,"published_at":33},"ai-agents-develop-self-replicating-malware-in-unsupervised-turf-war","AI Agents Develop Self-Replicating Malware in Unsupervised Turf War","2026-08-17T22:21:30.97+00:00",{"slug":35,"title":36,"published_at":37},"ai-guardrails-bypass-via-simple-authorization-claims","AI Guardrails Bypass via Simple Authorization Claims","2026-08-05T16:20:47.874+00:00",{"slug":39,"title":40,"published_at":41},"ai-jailbreaking-weaponized-as-offensive-attack-platform","AI Jailbreaking Weaponized as Offensive Attack Platform","2026-07-21T20:21:10.901+00:00",{"slug":43,"title":44,"published_at":45},"skillcloak-enables-malicious-ai-agent-skills-to-bypass-static-detection","SkillCloak Enables Malicious AI Agent Skills to Bypass Static Detection","2026-07-06T08:21:08.951+00:00",{"slug":47,"title":48,"published_at":49},"ai-agent-tool-descriptions-can-be-poisoned-to-leak-corporate-data","AI Agent Tool Descriptions Can Be Poisoned to Leak Corporate Data","2026-06-30T20:20:23.208+00:00",{"slug":51,"title":52,"published_at":53},"macos-malware-gaslight-weaponizes-ai-confusion-to-evade-analysis","macOS Malware 'Gaslight' Weaponizes AI Confusion to Evade Analysis","2026-06-25T18:20:42.484+00:00",[55,59,64,69,73,77],{"slug":56,"name":57,"problem":58,"stories":9},"ai-security","AI application security","Prompt injection, data leaks through AI tools and unapproved AI use get caught.",{"slug":60,"name":61,"problem":62,"stories":63},"threat-intelligence","Threat intelligence","Attack campaigns reach your sector, suppliers or software before your team hears about them.",3,{"slug":65,"name":66,"problem":67,"stories":68},"endpoint-detection-response","Endpoint detection and response (EDR and XDR)","Malware and hands-on-keyboard attackers on laptops and servers get spotted and stopped.",2,{"slug":70,"name":71,"problem":72,"stories":68},"identity-governance","Identity governance and single sign-on","People who leave or change roles lose access everywhere at once, and forgotten accounts stop piling up.",{"slug":74,"name":75,"problem":76,"stories":68},"security-awareness-training","Security awareness training and phishing simulation","People recognise and report attacks instead of falling for them.",{"slug":78,"name":79,"problem":80,"stories":81},"breach-attack-simulation","Breach and attack simulation","You only find out whether your security tools stop an attack when a real attacker tries.",1,[83,89,95,100,105,109,113,118],{"family":84,"item":85,"code":86,"title":87,"together":88},"nist-ai-rmf","overview","AI RMF 1.0","AI Risk Management Framework (AI RMF 1.0)",8,{"family":90,"item":91,"code":92,"title":93,"together":94},"iso","42001","ISO\u002FIEC 42001","ISO\u002FIEC 42001: Artificial intelligence management system",4,{"family":96,"item":97,"code":98,"title":99,"together":94},"cis","10","Control 10","Malware Defenses",{"family":101,"item":102,"code":103,"title":104,"together":63},"nist-800-53","si-10","SI-10","Information Input Validation",{"family":96,"item":106,"code":107,"title":108,"together":63},"8","Control 8","Audit Log Management",{"family":96,"item":110,"code":111,"title":112,"together":63},"2","Control 2","Inventory and Control of Software Assets",{"family":114,"item":115,"code":116,"title":117,"together":63},"gdpr","25","Article 25","Data protection by design and by default",{"family":101,"item":119,"code":120,"title":121,"together":63},"ac-6","AC-6","Least Privilege"]