Mistral AI Security Basics

Mistral AI Security Basics

Overview

Security best practices for Mistral AI API keys and access control.

Prerequisites

• Mistral AI SDK installed
• Understanding of environment variables
• Access to Mistral AI console

Instructions

Step 1: Configure Environment Variables

# .env (NEVER commit to git)
MISTRAL_API_KEY=your-api-key-here

# .gitignore - Add these lines
.env
.env.local
.env.*.local
*.pem
*.key
secrets/

Validate .gitignore:

# Check if .env would be committed
git check-ignore .env || echo "WARNING: .env is not ignored!"

# Check for exposed secrets in history
git log -p --all -S 'MISTRAL_API_KEY' -- '*.ts' '*.js' '*.py' '*.env'

Step 2: Secure Key Loading

// src/​config/​mistral.ts
import { z } from 'zod';

const envSchema = z.object({
  MISTRAL_API_KEY: z.string().min(20, 'Invalid API key format'),
  NODE_ENV: z.enum(['development', 'staging', 'production']).default('development'),
});

export function loadMistralConfig() {
  const result = envSchema.safeParse(process.env);

  if (!result.success) {
    console.error('Environment validation failed:', result.error.format());
    throw new Error('Missing or invalid MISTRAL_API_KEY');
  }

  return result.data;
}

// Usage
const config = loadMistralConfig();
const client = new Mistral({ apiKey: config.MISTRAL_API_KEY });

Step 3: Key Rotation Procedure

# 1. Generate new key in Mistral console
#    https://console.mistral.ai/​api-keys/

# 2. Test new key before rotation
MISTRAL_API_KEY_NEW="new-key-here"
curl -H "Authorization: Bearer ${MISTRAL_API_KEY_NEW}" \
  https://api.mistral.ai/​v1/​models

# 3. Update environment variable / secrets manager
export MISTRAL_API_KEY="${MISTRAL_API_KEY_NEW}"

# 4. Deploy with new key

# 5. Revoke old key in Mistral console

Step 4: Environment-Specific Keys

// src/​config/​environments.ts
interface MistralEnvConfig {
  apiKey: string;
  allowedModels: string[];
  maxTokensPerRequest: number;
}

const configs: Record<string, MistralEnvConfig> = {
  development: {
    apiKey: process.env.MISTRAL_API_KEY_DEV!,
    allowedModels: ['mistral-small-latest', 'mistral-large-latest'],
    maxTokensPerRequest: 1000,
  },
  production: {
    apiKey: process.env.MISTRAL_API_KEY_PROD!,
    allowedModels: ['mistral-small-latest'],
    maxTokensPerRequest: 4000,
  },
};

export function getEnvConfig(): MistralEnvConfig {
  const env = process.env.NODE_ENV || 'development';
  return configs[env] || configs.development;
}

Step 5: Request Validation & Sanitization

import { z } from 'zod';

const messageSchema = z.object({
  role: z.enum(['system', 'user', 'assistant']),
  content: z.string().max(100000), // Prevent huge payloads
});

const chatRequestSchema = z.object({
  model: z.enum(['mistral-small-latest', 'mistral-large-latest']),
  messages: z.array(messageSchema).min(1).max(100),
  temperature: z.number().min(0).max(1).optional(),
  maxTokens: z.number().min(1).max(4000).optional(),
});

function validateChatRequest(input: unknown) {
  return chatRequestSchema.parse(input);
}

// Usage in API endpoint
app.post('/​api/​chat', async (req, res) => {
  try {
    const validated = validateChatRequest(req.body);
    const response = await client.chat.complete(validated);
    res.json(response);
  } catch (error) {
    if (error instanceof z.ZodError) {
      res.status(400).json({ error: 'Invalid request', details: error.errors });
    } else {
      res.status(500).json({ error: 'Internal error' });
    }
  }
});

Step 6: Audit Logging

interface AuditEntry {
  timestamp: Date;
  action: string;
  userId?: string;
  model: string;
  tokensUsed?: number;
  success: boolean;
  errorCode?: string;
  ipAddress?: string;
}

async function auditLog(entry: Omit<AuditEntry, 'timestamp'>): Promise<void> {
  const log: AuditEntry = { ...entry, timestamp: new Date() };

  // Log to your audit system
  console.log('[AUDIT]', JSON.stringify(log));

  // Optional: send to logging service
  // await loggingService.record(log);
}

// Usage
async function chatWithAudit(
  userId: string,
  messages: Message[],
  ipAddress?: string
): Promise<string> {
  const start = Date.now();
  try {
    const response = await client.chat.complete({
      model: 'mistral-small-latest',
      messages,
    });

    await auditLog({
      action: 'chat.complete',
      userId,
      model: 'mistral-small-latest',
      tokensUsed: response.usage?.totalTokens,
      success: true,
      ipAddress,
    });

    return response.choices?.[0]?.message?.content ?? '';
  } catch (error: any) {
    await auditLog({
      action: 'chat.complete',
      userId,
      model: 'mistral-small-latest',
      success: false,
      errorCode: error.status?.toString(),
      ipAddress,
    });
    throw error;
  }
}

Security Checklist

• API key stored in environment variable (not code)
• .env files in .gitignore
• Different keys for dev/​staging/​prod
• Input validation on all requests
• Output sanitization before display
• Audit logging enabled
• Rate limiting implemented
• Error messages don't expose internals
• HTTPS enforced for all endpoints
• Regular key rotation scheduled

Output

• Secure API key storage
• Environment-specific access controls
• Input validation
• Audit logging enabled

Error Handling

Examples

Secrets Manager Integration (AWS)

import { SecretsManager } from '@aws-sdk/​client-secrets-manager';

const sm = new SecretsManager({ region: 'us-east-1' });

async function getMistralApiKey(): Promise<string> {
  const secret = await sm.getSecretValue({
    SecretId: 'mistral/​api-key',
  });
  return secret.SecretString!;
}

Secrets Manager Integration (GCP)

import { SecretManagerServiceClient } from '@google-cloud/​secret-manager';

const client = new SecretManagerServiceClient();

async function getMistralApiKey(): Promise<string> {
  const [version] = await client.accessSecretVersion({
    name: 'projects/​my-project/​secrets/​mistral-api-key/​versions/​latest',
  });
  return version.payload?.data?.toString()!;
}

Content Filtering

const BLOCKED_PATTERNS = [
  /\b(password|secret|key)\s*[:=]/​i,
  /\b\d{16}\b/, // Credit card numbers
  /\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z]{2,}\b/​i, // Emails
];

function containsSensitiveData(text: string): boolean {
  return BLOCKED_PATTERNS.some(pattern => pattern.test(text));
}

async function safeChatComplete(messages: Message[]): Promise<string> {
  // Check input for sensitive data
  for (const msg of messages) {
    if (containsSensitiveData(msg.content)) {
      throw new Error('Message contains potentially sensitive data');
    }
  }

  return client.chat.complete({ model: 'mistral-small-latest', messages });
}

Resources

• Mistral AI Console
• OWASP API Security
• Zod Documentation

Next Steps

For production deployment, see mistral-prod-checklist.