Skip to main content
Version: 6.1

AI Security: Jobs

The AI Security module includes 27 jobs based on the OWASP Top 10 for LLM Applications classification. The jobs run on a schedule and create an incident in Incident Manager when threat indicators are detected.

Intended audience: information security analysts and SOC engineers


Jobs List

CategoryJobDescription
LLM01Prompt injectionDetects prompt injection attempts in incoming requests by using a signature library that covers instruction overrides, role substitution, system prompt extraction, and jailbreaks (Incident example)
LLM03Unauthorized modification of training dataDetects training dataset files written, modified, or deleted by users who are not on the authorized users list
LLM03Unusual access to training dataDetects training dataset files read by users who are not on the authorized users list
LLM04Request context length limit exceededDetects requests whose input context exceeds the configured threshold (Incident example)
LLM04Generation time limit exceededDetects requests whose processing time exceeds the configured threshold, which may indicate a resource exhaustion attack
LLM04User token limit exceededCompares a user's token usage over 15 minutes with the average for the preceding 48 hours and triggers when usage exceeds the average by the configured factor
LLM04Total token limit exceededMonitors total token usage across all requests over 15 minutes against the configured threshold
LLM04Request rate threshold exceededCounts the number of requests per user per minute and compares it with the configured threshold
LLM05Unusual external agent connectionDetects outbound agent connections to destinations that are not on the allowed destinations list
LLM05/08Excessive permissions granted to agentsDetects dangerous combinations in permission configurations, including automatic approval without a denylist, overly broad command prefixes, and network access with automatic approval enabled
LLM05/08Deviation from the permission profileCompares the current agent permission configuration with the agent_permission_baseline lookup, including approval mode, network access, sandbox mode, policy, and allowlists
LLM05/08Agent privilege escalation attemptDetects failed agent actions and access denial messages such as permission denied, forbidden, and sandbox (Incident example)
LLM06Sensitive information disclosureAnalyzes model requests and responses for secrets, personal data, confidential business information, and web attack indicators by using a signature library (Incident example)
LLM06System prompt disclosure by the modelDetects characteristic markers of internal instructions and system prompts in model responses
LLM06Sensitive data leakage to external servicesDetects secrets, personal data, and confidential information sent in requests to external providers
LLM07Dangerous command in a model responseAnalyzes model responses for destructive, exfiltration, and privilege escalation commands (Incident example)
LLM07Dangerous process on an AI hostDetects processes launched with destructive, exfiltration, or malicious command lines
LLM07Dangerous action initiated by an agentDetects a dangerous command when its parent process is an AI agent runtime
LLM07Correlation: prompt injection to actionCorrelates a dangerous system action with an earlier prompt injection on the same host
LLM07Correlation: dangerous model response to actionCorrelates a dangerous system action with an earlier model response containing a dangerous command on the same host
LLM10Successful model repository access by an unusual userDetects access to model files by users who are not on the authorized users list
LLM10Multiple accesses to model filesCounts accesses to model weight files by user and host over a specified interval and triggers when the threshold is exceeded
LLM10Model artifact exfiltration by a processDetects data transfer commands such as scp, rsync, rclone, and curl when combined with access to model weight files
LLM10Outbound transfer of model weights to an external networkCorrelates large outbound connections with access to model files on the same host
LLM10API exfiltrationCompares a user's number of LLM API requests over 15 minutes with a 48-hour moving average; exceeding the average by the configured factor may indicate a model extraction attempt
InternalPopulate the agent_permission_baseline lookupCalculates current permission values for each agent daily and stores a baseline permission profile for the deviation detection rule
Important

Thresholds configured in the jobs can be adjusted for the target deployment.