RiskFramework.org
GitHub LinkedIn

Practical tools for better risk decisions.

Tools

CROSSWALK Mapping rev 1 Open for comment CC BY-SA 4.0 Updated 2026

Tool connections

How the tools on this site fit together. Each connection says what it links, which versions it was built on, and how far to trust it.

AI Risk Map™ × Four Faces of AI Risk™

The AI Risk Map lists the risk factors that come with your own use of AI, organized by who owns them. Four Faces sorts AI risk by cause. This mapping places every active AI Risk Map factor in one or two faces, so you can start from either tool and find the other's view.

The AI Risk Map covers faces 1 to 3, the faces about our own AI. Face 4 is outside its subject boundary by design, so that column holds no factors.

Get the mapping

  • Face per factor, IDs onlyJSON
  • AI Risk Map with a Four Faces columnCSV
Get the files on GitHub

In the AI Risk Map repository, under companions/four-faces/.

Domains by face

Each domain sits in the face where most of its factors fall. A dashed block is the part of a domain whose factors sit in another face, labelled with the categories they come from.

Mapping revision 1 · AI Risk Map v2.1 · Four Faces v1.2

1 · StrategyInnovate with AI

Are we making the right bets?

02Strategy & Value Realization
  • AIRM-193Our advantage erodes as competitors and customers approximate it with AI
  • AIRM-194AI agents disintermediate us from our customers
  • AIRM-026Failure to weigh the competitive risk of not adopting AI (inaction)
  • AIRM-020Lack of a coherent AI strategy aligned to objectives and risk appetite
  • AIRM-021AI adoption driven by hype rather than demonstrated business value
  • AIRM-022Inability to adapt the roadmap as foundation-model capabilities shift
  • AIRM-204Fear-driven AI bans push use into the shadows
  • AIRM-023Use cases chosen without a clear problem, success measure, or risk–benefit weighing
  • AIRM-024‘Pilot purgatory’ — pilots stalling on data readiness, evaluation difficulty, or inability to demonstrate value
  • AIRM-025Misjudged AI sourcing fork: train, fine-tune, RAG, or API
  • AIRM-027Unrealistic ROI expectations, including returns that need workflow redesign
  • AIRM-028Inability to attribute value actually delivered by AI investments
  • AIRM-029Hidden total cost of ownership (data, talent, integration) eroding returns
10Workforce, Human & Societal
  • AIRM-151Shortage of AI and AI-risk talent, competing against AI-native employers for scarce skills
  • AIRM-152Inability to retain specialists, who leave taking undocumented model and prompt knowledge with them
  • AIRM-195Leaders manage AI like a conventional IT rollout
  • AIRM-154Erosion of staff expertise through over-delegation to AI
  • AIRM-155Loss of institutional knowledge as AI replaces human processes
  • AIRM-157Change management underestimated as AI reshapes the work itself
  • AIRM-158Workforce resistance or low adoption driven by displacement fear or distrust of AI output
  • AIRM-159Labor-relations and morale impacts of AI deployment
  • AIRM-161Customers distrust or avoid AI-mediated services
11Financial & Resource
  • AIRM-165Poor AI cost forecasting as usage scales to production
  • AIRM-166AI's energy and water use unmeasured and unweighed
  • AIRM-167Stranded investment, including AI scaled too early or funded past failure
  • AIRM-168Procurement exposure where providers reprice tokens, tiers, or context windows unilaterally
  • AIRM-170Compute, GPU, or power scarcity constraining scaling
  • AIRM-172Concentration of compute supply among few providers
01Governance (part)Governance & oversight · Roles & accountability · Risk-management integration
  • AIRM-001Inadequate executive/board oversight of AI adoption and risk
  • AIRM-003Absence of an AI governance body or charter with a clear mandate
  • AIRM-004Unclear ownership and accountability (RACI) for AI risk decisions
  • AIRM-005No clear executive owner for AI adoption
  • AIRM-012AI risk not integrated into enterprise risk management (ERM)
  • AIRM-014No defined risk appetite or tolerance thresholds for AI
09Third-Party (part)Foundation-model dependence · Vendor due diligence · Lock-in & portability
  • AIRM-139Concentration risk from dependence on a few model providers
  • AIRM-140Model monoculture, including one model's blind spot reproduced across everything it produces
  • AIRM-141Provider deprecation or retirement of a model forcing unplanned migration
  • AIRM-143Vendor viability risk in an unusually volatile provider market with frequent pivots and failures
  • AIRM-145Vendor / model lock-in limiting portability and leverage
  • AIRM-146High switching cost from proprietary APIs and fine-tuning
  • AIRM-147Lack of exit provisions, where fine-tunes, embeddings, and prompt assets do not transfer between providers
2 · ExecutionBuild and run AI responsibly

Is it well built and well run?

01AI Governance & Accountability
  • AIRM-001Inadequate executive/board oversight of AI adoption and risk
  • AIRM-002Ungoverned ‘shadow AI’ adoption outside sanctioned channels
  • AIRM-003Absence of an AI governance body or charter with a clear mandate
  • AIRM-004Unclear ownership and accountability (RACI) for AI risk decisions
  • AIRM-005No clear executive owner for AI adoption
  • AIRM-006Accountability for agent decisions undefined, internally and across the AI stack
  • AIRM-007Segregation of duties collapses across AI config, operation, and review
  • AIRM-008Absence of an AI acceptable-use policy and model-approval gates
  • AIRM-009AI policies not kept current with evolving capabilities and regulation
  • AIRM-010Inconsistent standards as AI proliferates faster than governance
  • AIRM-011No agreed definition of ‘AI’, leaving policy and inventory scope undefined
  • AIRM-012AI risk not integrated into enterprise risk management (ERM)
  • AIRM-013Failure to conduct AI impact assessments for individuals and society
  • AIRM-014No defined risk appetite or tolerance thresholds for AI
  • AIRM-015Inability to audit AI: no complete inventory of models, agents, and connectors
  • AIRM-016Insufficient internal-audit competence to assess AI controls
  • AIRM-017Absence of independent assurance over AI control effectiveness
  • AIRM-018AI output relied on in financial-reporting controls without scope change
  • AIRM-019No way to identify which artifacts AI produced, so a model flaw cannot be scoped
03Data & Privacy
  • AIRM-030Unverified provenance or integrity of training, fine-tuning, or RAG data
  • AIRM-031Reliance on data scraped without lawful basis or clear rights
  • AIRM-032Inadequate data lineage to trace data through the AI pipeline
  • AIRM-033Poor data quality undermining model reliability
  • AIRM-034Inadequate or biased labeling introducing systematic error
  • AIRM-035Stale data, including grounding and retrieval sources that drift out of date behind a confident-sounding model
  • AIRM-036Personal data used for training without consent or purpose limitation
  • AIRM-037Inability to honor data-subject rights once data is in a model
  • AIRM-038Re-identification of individuals from aggregated or model-derived data
  • AIRM-039No guardrails keeping confidential or personal data out of model output
  • AIRM-040Unclassified sensitive data in AI, so handling controls never apply
  • AIRM-041Confidential data pasted into an approved AI tool, leaving despite working controls
  • AIRM-042Use of copyrighted or licensed material in training without rights
  • AIRM-043Contamination of datasets with third-party IP or trade secrets
  • AIRM-044Inability to prove you held rights to the training data used
  • AIRM-045Failure to preserve prompts, outputs, and agent traces subject to litigation hold or e-discovery
  • AIRM-046Deletion at source does not remove data absorbed into models
  • AIRM-047Persistence of sensitive data in embeddings, caches, or model weights
  • AIRM-048Failure to remove data across distributed / replicated AI stores
04Model Development & Performance
  • AIRM-049Inappropriate model/algorithm choice for the problem and risk context
  • AIRM-050Poorly documented or untracked model development
  • AIRM-051Insecure or unmanaged hyperparameters and training configuration
  • AIRM-052Hallucinated or meaning-changing output accepted as reliable
  • AIRM-053Validation data unrepresentative; real performance shows only with live users
  • AIRM-054Skipped or insufficient red-teaming, safety, and fairness testing
  • AIRM-055Absence of independent model validation / effective challenge
  • AIRM-056Model drift degrading performance as data / conditions change
  • AIRM-057Lack of robustness to edge cases and distribution shift
  • AIRM-058Small changes to the input produce very different output
  • AIRM-059Black-box models used in high-stakes decisions without interpretability
  • AIRM-060Inability to produce adverse-action / decision explanations
  • AIRM-061Over-reliance on post-hoc explanations that misrepresent the model
  • AIRM-062Non-reproducible training pipelines undermining validation and audit
  • AIRM-063Inadequate model documentation / model cards
  • AIRM-064Loss of traceability between model versions, data, and results
06Safety, Ethics & Responsible AI
  • AIRM-086Discriminatory or biased outputs causing disparate impact
  • AIRM-087Inequitable performance across languages, demographics, and disabilities
  • AIRM-088Bias amplification / homogenization from feedback and scale
  • AIRM-089Culturally inappropriate output as AI scales across regions
  • AIRM-090Generation of toxic, dangerous, or hateful content
  • AIRM-091Harmful, misleading, or defamatory output driving decisions
  • AIRM-092Inappropriate or unsafe AI output reaching minors or other vulnerable users
  • AIRM-199No guardrails on what customer-facing AI says in our name
  • AIRM-202AI makes false statements about real people or organizations
  • AIRM-093Failure to disclose to users that they are interacting with AI
  • AIRM-094Undisclosed synthetic media / deepfakes lacking provenance
  • AIRM-095Inadequate transparency to affected parties about AI's role
  • AIRM-192Our AI service changes under customers without notice, so their validations go stale
  • AIRM-098AI optimized for a proxy measure that diverges from users' interests
  • AIRM-099User over-trust / automation bias in AI outputs
  • AIRM-100Anthropomorphism and emotional dependence on conversational AI
  • AIRM-101Erosion of meaningful human control over decisions
  • AIRM-102Review capacity cannot scale to AI output volume, degrading oversight to rubber-stamping
07Operations & Resilience
  • AIRM-103Immature MLOps / fragile deployment pipelines
  • AIRM-104Inconsistent or unsanctioned (shadow) model deployment
  • AIRM-105Environment and config management gaps undermining reproducibility
  • AIRM-106Training-serving skew degrading live performance
  • AIRM-107AI tool-execution or test containment fails and AI reaches real systems
  • AIRM-108No monitoring, acted on, for drift, biased outcomes, or abuse
  • AIRM-109Inadequate logging of inputs, outputs, tool calls, agent decisions, and who authorized them
  • AIRM-110Non-determinism making anomalous behavior hard to define or detect
  • AIRM-111Sanctioned agent activity indistinguishable from agentic attack in telemetry
  • AIRM-112Inference latency or unavailability degrading dependent services
  • AIRM-113Inability to scale inference to peak demand
  • AIRM-115Immature AI incident response and escalation processes
  • AIRM-116Inability to investigate AI incidents due to non-determinism / poor logs
  • AIRM-117Inadequate content-moderation / trust-&-safety operations at scale
  • AIRM-118Defensive model refuses to analyze attack data mid-incident
  • AIRM-201No tested off-switch or fallback for a misbehaving AI feature
  • AIRM-119Vendor or model updates changing behavior, often silently or without notice
  • AIRM-120Inadequate versioning, change control, and tested rollback to a known-good model version
  • AIRM-121Unmanaged model decommissioning / sunset leaving orphaned dependencies
  • AIRM-122Uncontrolled continuous learning introducing drift
  • AIRM-123Controls re-verified less often than the AI system changes
  • AIRM-124Prompts and retrieval config changed outside change control
08Legal, Regulatory & Compliance
  • AIRM-125Unknowingly deploying a prohibited or high-risk AI use case
  • AIRM-126AI obligations not identified or built in before deployment
  • AIRM-127Inability to keep pace with fragmented, fast-changing AI regulation
  • AIRM-128User inability to contest or seek recourse on an AI-driven decision
  • AIRM-129Misclassified as deployer when fine-tuning makes you a provider
  • AIRM-200No way to find and make whole everyone a flawed AI decision affected
  • AIRM-130Unsettled liability for harm caused by AI decisions
  • AIRM-131AI-generated output that closely reproduces or imitates others' copyrighted work or trademarks
  • AIRM-132Uncertain ownership / copyrightability of AI-generated content
  • AIRM-133Vendor contracts lacking AI terms: training use, output ownership, change notice
  • AIRM-134Use of open-weight or licensed models outside their license terms
  • AIRM-135Failure to update contracts as AI use and regulation evolve
  • AIRM-136Data crossing borders through AI services, including hidden routing
  • AIRM-137Conflicting regional AI requirements the organization cannot reconcile for a single system (externally imposed)
  • AIRM-138Inability to evidence and audit AI compliance across jurisdictions (internal evidence burden)
09Third-Party & Model Supply Chain
  • AIRM-140Model monoculture, including one model's blind spot reproduced across everything it produces
  • AIRM-142Vendor diligence missing: are prompts retained, trained on, or shared?
  • AIRM-188No way to verify a vendor's 'AI-powered' claim before buying it
  • AIRM-144Accountability for a vendor's AI failures staying with the organization, whatever the contract says
  • AIRM-149Open-source AI dependencies abandoned by their maintainers
  • AIRM-150Opaque sub-processors, including undisclosed models and embedded AI features
  • AIRM-189Models deployed without verifying integrity or lineage
  • AIRM-191Safety training stripped or eroded in a self-hosted model
12Autonomy & Delegated Action (Agentic AI)
  • AIRM-173Agent pursues unintended goals or misgeneralizes its objective
  • AIRM-174Agent pursues an authorized objective through unauthorized means
  • AIRM-175Excessive autonomy / agency beyond intended mandate
  • AIRM-176No working interrupt or kill switch for an agent mid-task
  • AIRM-206No human approval before high-consequence agent actions
  • AIRM-203Agents that rewrite their own instructions or tools
  • AIRM-177Agent granted a broader set of invokable tools / functions than its task requires
  • AIRM-179Third-party tool/connector servers as an unvetted execution surface
  • AIRM-182Persistence of corrupted or sensitive data in agent memory
  • AIRM-183Cascading failures or loops across interacting agents
  • AIRM-184Misplaced inter-agent trust or agent collusion
  • AIRM-185Emergent, unpredictable behavior in multi-agent systems
  • AIRM-186Agents able to take irreversible real-world actions
  • AIRM-187Accountability gap for harm caused by autonomous action
  • AIRM-190Third parties can't tell your agent acted, so no one can warn you
05Security (part)Output & integration security · AI access & identity
  • AIRM-075Unsanitized output reaching downstream interpreters
  • AIRM-077Over-trusted output triggering unsafe automated actions
  • AIRM-085Retrieval serving content beyond the user’s own entitlements
10Workforce (part)Skills & talent · Deskilling & loss of institutional knowledge · Societal & reputational
  • AIRM-152Inability to retain specialists, who leave taking undocumented model and prompt knowledge with them
  • AIRM-153Inadequate AI literacy across the workforce
  • AIRM-156Capability atrophy — inability to operate if AI is unavailable
  • AIRM-160Reputational damage amplified when AI fails in public
  • AIRM-162AI decisions affecting people who never chose to use the system
  • AIRM-163AI-washing: overstating capability to customers, investors, or regulators
11Financial (part)Cost & compute economics · Investment / procurement
  • AIRM-164Uncontrolled AI consumption driving up inference spend
  • AIRM-166AI's energy and water use unmeasured and unweighed
  • AIRM-169Insurance excludes AI losses, or defines 'user' as human only
3 · SubversionPrevent abuse of our AI

Can it be turned against us?

05Security & Adversarial Threats
  • AIRM-065Training-data poisoning that degrades or backdoors model behavior
  • AIRM-066Tampering with RAG / knowledge-base sources consumed at inference
  • AIRM-067Label-flipping or adversarial dataset partitioning
  • AIRM-068Model, reasoning, or insight extraction through use
  • AIRM-069Membership-inference or model-inversion exposing training data
  • AIRM-070Backdoored / trojaned pre-trained model from a public hub
  • AIRM-071Direct prompt injection altering behavior or bypassing guardrails
  • AIRM-072Indirect / cross-domain prompt injection via retrieved content
  • AIRM-073Jailbreaks defeating safety controls and restrictions
  • AIRM-074System-prompt leakage exposing operational instructions
  • AIRM-197Inputs crafted to evade AI detection models
  • AIRM-205Untrusted content treated as instructions
  • AIRM-075Unsanitized output reaching downstream interpreters
  • AIRM-076Interception or manipulation of the inference output stream
  • AIRM-077Over-trusted output triggering unsafe automated actions
  • AIRM-078Compromised or malicious model from the AI supply chain
  • AIRM-079Known vulnerabilities in ML dependencies not patched in time
  • AIRM-080Tampering with model artifacts, prompts, or configuration
  • AIRM-081Model-invented dependencies registered by attackers
  • AIRM-082Weak access controls and usage limits on AI endpoints
  • AIRM-083Over-privileged service accounts and keys for AI components
  • AIRM-084Credential sprawl as AI tools and agent frameworks outpace governance
  • AIRM-085Retrieval serving content beyond the user’s own entitlements
  • AIRM-196Insiders use AI agents to reach beyond their own access
01Governance (part)Roles & accountability · Audit & assurance
  • AIRM-007Segregation of duties collapses across AI config, operation, and review
  • AIRM-015Inability to audit AI: no complete inventory of models, agents, and connectors
03Data & Privacy (part)Privacy & consent
  • AIRM-038Re-identification of individuals from aggregated or model-derived data
04Model Development (part)Accuracy & validation
  • AIRM-054Skipped or insufficient red-teaming, safety, and fairness testing
06Safety & Ethics (part)Harmful / inaccurate output · Misuse & dual-use
  • AIRM-199No guardrails on what customer-facing AI says in our name
  • AIRM-096Misuse of AI for disinformation, fraud, or social engineering
  • AIRM-097Model gives meaningful help toward weapons or cyberattack capability
07Operations (part)MLOps & deployment · Monitoring & observability · Availability & scalability · Incident management · Change & lifecycle management
  • AIRM-107AI tool-execution or test containment fails and AI reaches real systems
  • AIRM-108No monitoring, acted on, for drift, biased outcomes, or abuse
  • AIRM-109Inadequate logging of inputs, outputs, tool calls, agent decisions, and who authorized them
  • AIRM-110Non-determinism making anomalous behavior hard to define or detect
  • AIRM-111Sanctioned agent activity indistinguishable from agentic attack in telemetry
  • AIRM-114Denial-of-wallet: attackers run up AI compute costs
  • AIRM-115Immature AI incident response and escalation processes
  • AIRM-116Inability to investigate AI incidents due to non-determinism / poor logs
  • AIRM-117Inadequate content-moderation / trust-&-safety operations at scale
  • AIRM-201No tested off-switch or fallback for a misbehaving AI feature
  • AIRM-123Controls re-verified less often than the AI system changes
08Legal & Compliance (part)Contractual & licensing
  • AIRM-198Customers and partners feeding our data into their AI, with no terms or visibility
09Third-Party (part)Open-source component risk
  • AIRM-148Malicious or compromised open-source models / components
  • AIRM-189Models deployed without verifying integrity or lineage
10Workforce (part)Societal & reputational
  • AIRM-160Reputational damage amplified when AI fails in public
11Financial (part)Investment / procurement
  • AIRM-169Insurance excludes AI losses, or defines 'user' as human only
12Autonomy (part)Autonomy, alignment & oversight · Tool access & authorization · Memory & context integrity · Multi-agent orchestration · Action integrity & traceability
  • AIRM-176No working interrupt or kill switch for an agent mid-task
  • AIRM-206No human approval before high-consequence agent actions
  • AIRM-177Agent granted a broader set of invokable tools / functions than its task requires
  • AIRM-178Agent identity misuse, impersonation, or delegated-credential abuse
  • AIRM-179Third-party tool/connector servers as an unvetted execution surface
  • AIRM-180Memory poisoning corrupting agent state across sessions
  • AIRM-181Context-window manipulation altering agent behavior
  • AIRM-184Misplaced inter-agent trust or agent collusion
  • AIRM-186Agents able to take irreversible real-world actions
4 · AttackDefend against their AI

Is theirs aimed at us?

Not on the AI Risk Map, by design

Deepfake impersonation, AI-enabled fraud and phishing, autonomous attack agents, frontier capability with the guardrails stripped.

Managed on the threat register.

Most of the domain sits in this facePart of a domain whose other factors sit in another faceA factor can sit in two faces. Open any block to see its factors.

How the mapping changes

The mapping has its own revision number, pinned to the AI Risk Map version and the Four Faces version it was made under. A change to the mapping alone is a new revision of this page and the companion files, not a new release of the AI Risk Map. Every AI Risk Map release re-checks it.

Connects

AI Risk MapMap · v2.1 · CC BY-SA 4.0 See the map →

With

Four Faces of AI RiskFramework · v1.2 · CC BY 4.0 See the faces →

Discuss

This mapping is open for comment. If a factor sits in the wrong face, say so in a GitHub issue on the AI Risk Map repository.