← All posts

Problem-Aware

249 posts on problem-aware.

Shadow AI Policy Template: What a Defensible Internal Policy Actually Contains

A shadow AI policy is the document a regulator reads first when something goes wrong. Most copy-paste templates fail because they list rules without the enforcement architecture behind them. This piece walks through the seven sections a defensible policy contains, the enforcement architecture each section assumes, and where most published templates fall short of what an EU AI Act reviewer or a HIPAA auditor will actually accept.

shadow-aiai-governancepolicyai-securitycomplianceaudit
Read post →

How to Write an AI Usage Policy That Holds Up Under Audit

An AI usage policy that survives a regulatory review covers data classification, identity binding, sanctioned tools, prompt content rules, audit retention, and incident handling. The pattern that fails most often is a policy written for HR distribution that the security team cannot demonstrate compliance with. This piece walks through the eight sections every policy needs, the enforcement layer the policy depends on, and the audit evidence the policy has to produce.

ai-usage-policyai-governancecomplianceshadow-aipolicy-templateaudit
Read post →

Enterprise AI Usage Policy Template: The Eight Sections That Survive Both Workforce Adoption and Regulatory Review

An AI usage policy that an enterprise can actually enforce contains eight sections: scope and definitions, sanctioned tools list, data classification rules, role-based permissions, the disclosure obligation, the inspection and monitoring statement, the incident reporting path, and the policy version history. A policy without inspection architecture behind it leaves the enterprise with a written commitment the workforce can ignore. The eight sections align with the EU AI Act Article 26 deployer obligations and the Article 12 record-keeping mandate.

ai-governanceai-policycomplianceshadow-aiauditai-security
Read post →

AI Usage Policy Examples: Six Working Templates by Industry

Working AI usage policy examples have to match the regulatory regime they live under. The healthcare policy turns on PHI and the BAA. The financial services policy turns on MNPI and DORA. The SaaS policy turns on customer data and the EU AI Act deployer obligations. This piece walks through six industry-calibrated policy examples, the specific clauses that distinguish them, and the enforcement layer all six share.

ai-usage-policypolicy-templatecomplianceai-governanceshadow-aiindustry
Read post →

Best-of-N Jailbreaking Turns Repeated Requests Into One Attack

Best-of-N jailbreaking repeatedly perturbs the same harmful objective and sends each variation as a separate model request until one succeeds. The attack turns request volume and model variance into an advantage. Production defenses need identity-bound attempt budgets, objective-level correlation, response inspection, and records that preserve the whole request series rather than treating each prompt as an isolated event.

ai-securityllm-securityprompt-injectioninline-enforcementpolicy-enforcement
Read post →

CISA Advisory Exposes AI Model Distillation Transfer Stations

The September 2026 CISA advisory, published with NSA and FBI, describes transfer stations that resell access to frontier models while obscuring origin and spreading requests across accounts and providers. I explain the distributed traffic mechanism and the enterprise egress evidence it changes, along with the exact boundary of an identity-aware HTTP policy gateway.

ai-securitycybersecurityllm-securityshadow-aiidentity-and-authorizationpolicy-enforcement
Read post →

Reward Hacking at the API Boundary: Detect Symptoms Without Claiming a Cure

Reward hacking occurs when a system optimizes a proxy in ways that defeat the intended goal. An HTTP gateway cannot inspect internal objectives or retrain a model, but it can preserve authenticated request sequences, model versions, policy state, outputs, and escalation evidence for observable post-deployment symptoms.

ai-securitymodel-riskincident-responseauditllm-security
Read post →

AI Bug Bounty: Scope, Evidence, Rewards, and Runtime Handoff

An AI bug bounty pays independent researchers for valid findings against a declared AI system scope. A workable program defines eligible prompts, models, APIs, retrieval paths, identities, data, and tools, then specifies safe-harbor rules, evidence, severity, duplicate handling, reward decisions, remediation, retesting, and the handoff into production controls.

ai-securityllm-securityprompt-injectionauditpolicy-enforcementcybersecurity
Read post →

AI-Generated Malware Needs an Abuse-Control Test Pack

AI-generated malware controls need to distinguish prohibited malicious-code requests from authorized security work, then preserve the route, caller, prompt and output classifications, verdict, escalation, and evidence. Anthropic’s September 2026 threat report describes an actor using AI agents to monitor detections and rebuild malware after detection. This guide turns that failure pattern into a non-operational abuse-control test pack with explicit ownership boundaries.

ai-securityllm-securitycybersecuritypolicy-enforcementforensic-auditai-governance
Read post →

AI-Generated Code Vulnerabilities Need a Production Acceptance Gate

AI-generated code can be syntactically plausible while carrying security flaws and dependency risk, exposing secrets, or adding behavior the prompt never requested. The control is a production acceptance gate that joins routed model-call evidence to human review, tests, SAST, dependency and secret scans, exceptions, merge state, and release evidence. Each system owns one part of the decision.

ai-securitydevsecopsllm-securitypolicy-enforcementaudit
Read post →

MCP Confused Deputy: Why the Server Acting on the User Is the Wrong Principal

The confused deputy attack describes the case where a privileged intermediary acts on behalf of a less-privileged caller and ends up doing things the caller could not have done directly. In the Model Context Protocol (MCP) world, the confused deputy lives in the MCP server. The MCP server holds credentials for upstream tools and acts on behalf of an LLM client. When the client identity is not propagated to the upstream calls, the upstream services see the MCP server, not the user, and authorization decisions get made against the wrong principal. This article walks the attack pattern, the architectural cause, the controls a policy gateway enforces at the MCP boundary, and the operational checklist.

mcpconfused-deputyauthorizationagentic-aipolicy-enforcementsecurity
Read post →

MCP Tool Poisoning Prevention: Gateway Controls for the Model Context Protocol Surface

Model Context Protocol tool poisoning is the agentic analog to supply-chain compromise. An MCP server presents a set of tools to an agent host; an attacker who controls the MCP server (or the tool definitions an MCP server advertises) can change what the tools do, what they return, or what parameters they accept. The agent loop calls the tool in good faith and the actions executed against downstream systems are the attacker'"'"'s. The prevention surface splits across MCP server selection, tool-definition pinning, and per-decision authorization at the agent-tool boundary. This article walks through the MCP poisoning patterns and the gateway controls that contain them.

mcpai-securityagentic-aitool-permissionssupply-chainai-gateway
Read post →