What it is
Repudiation is the "R" in STRIDE. In its classic form, it covers actors denying actions because no log proved otherwise. For AI agents the surface is broader: an attacker may flood the logging system to bury the relevant events; a misconfigured agent may not log tool arguments, only tool names; an agent may take an action under a shared service account so the actual requester cannot be traced; logs may exist but be unstructured, uncorrelated, or deleted on a 7-day rotation. The result is the same — when something goes wrong, you cannot prove what happened or who triggered it.
Why it matters
Audit-grade traceability is increasingly a regulatory requirement (the EU AI Act, sector rules in finance and healthcare, internal compliance for any SOC 2 or ISO 27001 environment). It is also what makes incident response possible at all — without it, you cannot tell whether an agent's bad action was a bug, an attack, or a user error. For Salesforce customers, agent actions touching customer data must trace back to a specific user, agent, prompt, and tool call. Treat logging as part of the agent, not as an afterthought.
Key components
- Insufficient logging — actions taken without recording the inputs that caused them
- Shared-identity actions — agents acting under service accounts that mask the real requester
- Log flooding — attackers generating noise that buries the signal
- Retention gaps — logs deleted before incidents are detected
- Mitigation — structured logs, end-to-end correlation IDs, tamper-evident storage, retention aligned to detection time
How it connects
In Agentforce, every agent action — whether it updates a record, fires an Apex action, or calls an external API — should be tied to a named user and a traceable prompt chain through Salesforce's audit trail and Event Monitoring. If your agents run under a single integration user or skip structured logging, you've already created a repudiation gap before the first incident occurs.
Good to know
Salesforce Event Monitoring logs API calls and record changes, but it won't automatically capture what an agent was asked to do or why — you need to explicitly design prompts, tool calls, and outcomes into your logging strategy from day one.
Related terms
Agent Governance
The policies, controls, and monitoring systems that ensure AI agents operate safely, compliantly, and within business-approved boundaries.
Agent Observability
The practice of inspecting, debugging, and understanding AI agent behavior at runtime by consuming agent telemetry — traces, metrics, logs, and events — through dashboards, alerts, and evaluation tools.
MELT (Metrics, Events, Logs, Traces)
The four-pillar framework for organizing observability data: Metrics (numeric signals), Events (state transitions), Logs (per-step context), and Traces (end-to-end request chains).
STRIDE
A six-category threat-modeling framework — Spoofing, Tampering, Repudiation, Information disclosure, Denial of service, Elevation of privilege — adapted from classic software security to AI agents.
Threat Modeling (AI Agents)
A structured way to identify what could go wrong with an AI agent before it does — mapping attackers, attack surfaces, and impacts so you can build defenses in.
