3. Permissions by design: Bind instruments to duties, to not fashions
A typical anti-pattern is to provide the mannequin a long-lived credential and hope prompts hold it well mannered. SAIF and NIST argue the alternative: credentials and scopes ought to be sure to instruments and duties, rotated recurrently, and auditable. Brokers then request narrowly scoped capabilities by means of these instruments.
In follow, that appears like: “finance-ops-agent might learn, however not write, sure ledgers with out CFO approval.”
The CEO query: Can we revoke a selected functionality from an agent with out re-architecting the entire system?
Management knowledge and habits
These steps gate inputs, outputs, and constrain habits.
4. Inputs, reminiscence, and RAG: Deal with exterior content material as hostile till confirmed in any other case
Most agent incidents begin with sneaky knowledge: a poisoned net web page, PDF, electronic mail, or repository that smuggles adversarial directions into the system. OWASP’s prompt-injection cheat sheet and OpenAI’s personal steering each insist on strict separation of system directions from consumer content material and on treating unvetted retrieval sources as untrusted.
Operationally, gate earlier than something enters retrieval or long-term reminiscence: new sources are reviewed, tagged, and onboarded; persistent reminiscence is disabled when untrusted context is current; provenance is hooked up to every chunk.
The CEO query: Can we enumerate each exterior content material supply our brokers be taught from, and who authorized them?
5. Output dealing with and rendering: Nothing executes “simply because the mannequin mentioned so”
Within the Anthropic case, AI-generated exploit code and credential dumps flowed straight into motion. Any output that may trigger a aspect impact wants a validator between the agent and the true world. OWASP’s insecure output dealing with class is express on this level, as are browser safety greatest practices round origin boundaries.
