Assess · Risk Assessment
AvailableA risk score you can explain to your board — and your auditor.
FlashTower's Risk Engine evaluates your answers against a versioned library of deterministic rules. Each rule has a permanent ID, a severity, a weight, a plain-language condition, an explanation, remediation steps, evidence requirements, and framework mappings. The same answers always produce the same result.
Details
Why deterministic
Security decisions need to be reproducible. Large language models are useful for explaining findings, but they should not decide your score. FlashTower never lets an AI model determine findings or scores; AI is used only, and optionally, to write narrative around deterministic results.
How findings are weighted
Each severity has a base point value, multiplied by the rule's weight. When the same rule applies to several AI systems, each additional occurrence counts for half the previous one, so one systemic issue does not overwhelm the score.
Versioned and auditable
Every assessment records the ruleset version and scoring version it was evaluated with. When rules change, their version number changes too, so past reports remain reproducible.
Methodology
How the score is calculated.
Scoring version 1.0.0, ruleset 2026.09.1. Higher scores mean more risk.
1. Points per finding
Each finding contributes its severity's base points multiplied by the rule's weight (0.5–1.5).
| Severity | Base points |
|---|---|
| Critical | 25 |
| High | 12 |
| Medium | 5 |
| Low | 2 |
| Info | 0 |
2. Repeat decay
When one rule applies to several AI systems, the n-th occurrence counts 0.5n−1 of full value. A single systemic issue is visible without overwhelming the score.
3. Saturating score
score = round(100 × (1 − e^(−points / 60)))
The curve keeps the score between 0 and 100. If any Critical finding exists, the score is raised to at least 60.
4. Rating bands
- Low 0–19
- Moderate 20–39
- Elevated 40–59
- High 60–79
- Critical 80–100
The ruleset
Every rule currently in the FlashTower Risk Engine. Rule IDs are permanent; changes increment the rule's version.
| Rule | Title | Category | Severity | Weight |
|---|---|---|---|---|
| FT-GOV-001 v1 | No inventory of AI tools and agents | Governance | High | 1.0 |
| FT-GOV-002 v1 | AI inventory is incomplete | Governance | Medium | 1.0 |
| FT-GOV-003 v1 | No named owner accountable for AI risk | Governance | Medium | 0.9 |
| FT-POL-001 v1 | No AI acceptable-use policy | Policies | High | 1.0 |
| FT-POL-002 v1 | AI use policy is still in draft | Policies | Medium | 0.8 |
| FT-POL-003 v1 | Employees have not acknowledged the AI policy | Policies | Low | 1.0 |
| FT-GOV-004 v1 | No AI security awareness training | Governance | Medium | 0.8 |
| FT-GOV-005 v1 | AI training reaches only part of the workforce | Governance | Low | 0.8 |
| FT-VEN-001 v1 | AI vendors are not reviewed before handling regulated data | Vendor Risk | High | 1.0 |
| FT-VEN-002 v1 | AI vendor review process is missing or partial | Vendor Risk | Medium | 0.9 |
| FT-IDN-001 v1 | No controls on unapproved (shadow) AI tools | Identity | High | 1.0 |
| FT-AUT-001 v1 | AI tools do not require multi-factor authentication | Authentication | High | 1.0 |
| FT-AUT-002 v1 | MFA is not enforced on all AI tools | Authentication | Medium | 0.8 |
| FT-AUT-003 v1 | AI tools are not connected to single sign-on | Authentication | Medium | 0.8 |
| FT-PRM-001 v1 | AI access is not periodically reviewed | Permissions | Medium | 1.0 |
| FT-LOG-001 v1 | No centralized logging while AI systems can take actions | Logging | High | 1.0 |
| FT-LOG-002 v1 | AI activity is not centrally logged | Logging | Medium | 0.9 |
| FT-IRP-001 v1 | Incident response does not cover AI | Incident Response | High | 0.9 |
| FT-IRP-002 v1 | Incident response only partially covers AI | Incident Response | Medium | 0.8 |
| FT-DAT-001 v1 | AI data retention not configured for regulated data | Data Exposure | Medium | 1.0 |
| FT-GOV-006 v1 | No AI systems were listed in this assessment | Governance | Info | 1.0 |
| FT-GOV-007 v1 | Significant visibility gaps | Governance | Info | 1.0 |
| FT-DAT-010 v1 | AI system accesses health information without a BAA | Sensitive Data | Critical | 1.2 |
| FT-DAT-011 v1 | Highly sensitive data in a consumer AI account | Sensitive Data | Critical | 1.1 |
| FT-DAT-012 v1 | Sensitive business data in a consumer AI account | Sensitive Data | High | 1.0 |
| FT-DAT-013 v1 | Sensitive data processed without a data processing agreement | Vendor Risk | Medium | 1.0 |
| FT-DAT-014 v1 | No confirmation vendor does not train on your data | Model Risk | Medium | 0.9 |
| FT-DAT-015 v1 | Payment card data reachable by AI system | Compliance | High | 0.9 |
| FT-CRD-001 v1 | Credentials embedded in code, configuration, or prompts | Credentials & Secrets | Critical | 1.1 |
| FT-CRD-002 v1 | AI system uses a shared key or shared account | Credentials & Secrets | High | 1.0 |
| FT-CRD-003 v1 | AI system acts with an employee's personal credentials | Identity | High | 1.0 |
| FT-CRD-004 v1 | Unknown how a high-impact AI system authenticates | Credentials & Secrets | Medium | 0.9 |
| FT-CRD-005 v1 | AI system can read passwords, keys, or secrets | Credentials & Secrets | High | 1.0 |
| FT-AGT-001 v1 | Unrestricted autonomous agent with high-impact permissions | Agent Risk | Critical | 1.2 |
| FT-AGT-002 v1 | Autonomous agent with high-impact permissions is not logged | Human Oversight | High | 1.0 |
| FT-AGT-003 v1 | Unknown whether a person approves high-impact AI actions | Human Oversight | High | 0.9 |
| FT-EXE-001 v1 | AI can execute code against production or cloud infrastructure without approval | Code Execution | Critical | 1.2 |
| FT-PRD-001 v1 | AI system has unsupervised production or cloud access | Production Access | High | 1.0 |
| FT-EXE-002 v1 | Human-approved AI code execution with production reach | Code Execution | Medium | 0.8 |
| FT-FIN-001 v1 | AI can move money without per-action approval | Financial Authority | Critical | 1.2 |
| FT-COM-001 v1 | AI sends or publishes externally without review | External Communications | High | 1.0 |
| FT-PRM-010 v1 | AI can change user permissions without approval | Authorization | Critical | 1.2 |
| FT-PRM-011 v1 | AI can change user permissions (with approval) | Authorization | Medium | 1.0 |
| FT-PRM-012 v1 | AI can delete data without approval | Data Exposure | High | 0.9 |
| FT-PRM-013 v1 | Broad permission scope | Permissions | Medium | 0.9 |
| FT-CDE-001 v1 | AI can push code without review | Agent Risk | High | 0.9 |
| FT-3PI-001 v1 | Prompt-injection exposure: web access combined with sensitive data or actions | Third-Party Integrations | High | 1.0 |
| FT-LOG-010 v1 | AI system with sensitive data is not logged | Logging | Medium | 0.9 |
| FT-GOV-010 v1 | Unapproved AI tool handles sensitive data | Governance | High | 1.0 |
| FT-GOV-011 v1 | Unapproved AI tool in use | Governance | Medium | 0.8 |
| FT-GOV-012 v1 | AI system has no named owner | Governance | Low | 1.0 |
| FT-VEN-010 v1 | Vendor security review not completed for sensitive-data AI system | Vendor Risk | Medium | 0.8 |
| FT-AUT-010 v1 | Team-wide use of consumer AI accounts | Identity | Medium | 0.8 |
See what your AI can access.
Run the free FlashTower Scan. Get your AI risk score and every finding in about 10 minutes — no account or system access required.