> ## Documentation Index
> Fetch the complete documentation index at: https://docs.alpha.ac/llms.txt
> Use this file to discover all available pages before exploring further.

# Rating Profiles

> Developer, Deployer, and Hybrid profiles determine how dimension weights are applied.

## Why Profiles Matter

A frontier model developer and a bank deploying third-party AI face fundamentally different governance challenges. Applying identical weights would penalize deployers for not publishing safety research and developers for not having vendor governance programs.

Alpha uses three profiles within the same 12-dimension framework. The dimensions are the same; the weights reflect what matters most for each company type.

## Profile Definitions

### Developer Profile

Companies that build, train, and distribute AI models.

**Examples:** Anthropic, OpenAI, Google DeepMind, Meta AI, xAI, Stability AI, Mistral

**Emphasis:** Safety (D2: 12.3%), Data Ethics (D4: 9.8%), Cybersecurity (D9: 10.7%), Corrigibility (D12: 10.0%)

### Deployer Profile

Companies that integrate, deploy, and operate AI systems built by others.

**Examples:** JPMorgan, UnitedHealth, Walmart, most Fortune 500 companies

**Emphasis:** Governance Infrastructure (D5: 11.4%), Leadership (D1: 9.8%), Fairness (D6: 8.2%), Supply Chain (D10: 8.2%)

### Hybrid Profile

Companies that both build and deploy AI systems.

**Examples:** Apple, Amazon, Microsoft, Google (cloud + models), Tesla

**Emphasis:** Balanced weights with slight emphasis on Safety (D2: 9.8%) and Governance Infrastructure (D5: 9.8%)

## Profile Assignment

Profile assignment is determined during Phase 1 (Evidence Collection) based on:

1. **Primary business activity:** Does the company develop foundation models or primarily deploy third-party AI?
2. **AI materiality:** How material is AI to the company's core business and risk profile?
3. **Model distribution:** Does the company distribute models to external users or deploy internally only?

The assigned profile is always disclosed in the rating output. Hybrid is the default when classification is ambiguous.

## Weight Comparison

| Dimension | Developer | Deployer | Hybrid |
| - | - | - | - |
| D1: Leadership | 8.2% | 9.8% | 9.0% |
| D2: Safety | **12.3%** | 6.6% | 9.8% |
| D3: Transparency | 8.2% | 8.2% | 8.2% |
| D4: Data Ethics | **9.8%** | 6.6% | 8.2% |
| D5: Governance Infra | 8.2% | **11.4%** | 9.8% |
| D6: Fairness | 4.9% | **8.2%** | 6.6% |
| D7: Workforce & Society | 4.9% | **8.2%** | 6.6% |
| D8: Regulatory | 8.2% | 8.2% | 8.2% |
| D9: Cybersecurity | **10.7%** | 6.6% | 8.2% |
| D10: Supply Chain | 6.6% | **8.2%** | 7.4% |
| D11: Stakeholder Impact | 8.0% | 8.0% | 8.0% |
| D12: Corrigibility | 10.0% | 10.0% | 10.0% |

D11 and D12 carry equal weight across all profiles: these are universal governance requirements regardless of whether a company builds or deploys AI.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.