OpenAI 推出全新企业隐私保护机制以对抗 Anthropic

import os
import requests

# 配置您的统一 n1n.ai API 凭证
N1N_API_KEY = os.getenv("N1N_API_KEY")
N1N_API_URL = "https://api.n1n.ai/v1/chat/completions"

def send_secure_llm_request(prompt: str, model_provider: str = "openai") -> str:
    """
    通过 n1n.ai 安全网关向指定的 LLM 提供商发送请求
    """
    headers = {
        "Authorization": f"Bearer {N1N_API_KEY}",
        "Content-Type": "application/json",
        # 传递自定义隐私控制头，声明零数据保留要求
        "X-Data-Retention-Policy": "zero-retention"
    }

    # 映射对应的模型名称
    model_name = "gpt-4o" if model_provider == "openai" else "claude-3-5-sonnet"

    payload = {
        "model": model_name,
        "messages": [
            {
                "role": "system",
                "content": "You are a secure enterprise assistant. Do not log or cache this session."
            },
            {
                "role": "user",
                "content": prompt
            }
        ],
        "temperature": 0.1
    }

    try:
        response = requests.post(N1N_API_URL, json=payload, headers=headers, timeout=30)
        response.raise_for_status()
        return response.json()["choices"][0]["message"]["content"]
    except requests.exceptions.RequestException as e:
        print(f"API 连接异常: {e}")
        # 发生故障时，自动切换至备用服务商
        if model_provider == "openai":
            print("正在通过 n1n.ai 自动切换至 Anthropic Claude 备用路由...")
            return send_secure_llm_request(prompt, model_provider="anthropic")
        raise e

# 示例调用
test_prompt = "分析以下敏感财务数据是否存在合规风险: [数据已脱敏]"
result = send_secure_llm_request(test_prompt, model_provider="openai")
print(result)

import re

class PIIMasker:
    def __init__(self):
        # 定义常见的 PII 正则表达式模式
        self.patterns = {
            "EMAIL": r"[a-zA-Z0-9_.+-]+@[a-zA-Z0-9-]+\.[a-zA-Z0-9-.]+",
            "IP_ADDRESS": r"\b(?:[0-9]{1,3}\.){3}[0-9]{1,3}\b",
            "CREDIT_CARD": r"\b(?:\d[ -]*?){13,16}\b"
        }
        self.vault = {}

    def mask(self, text: str) -> str:
        masked_text = text
        for entity_type, pattern in self.patterns.items():
            matches = re.findall(pattern, masked_text)
            for i, match in enumerate(matches):
                placeholder = f"__MASKED_{entity_type}_{i}__"
                self.vault[placeholder] = match
                masked_text = masked_text.replace(match, placeholder)
        return masked_text

    def unmask(self, text: str) -> str:
        unmasked_text = text
        for placeholder, original_value in self.vault.items():
            unmasked_text = unmasked_text.replace(placeholder, original_value)
        return unmasked_text

# 将脱敏逻辑与 API 请求相结合
masker = PIIMasker()
raw_prompt = "请发送密码重置链接至 [email protected] 并测试 IP 192.168.1.105 的连通性。"

# 第一步：在本地对输入内容进行脱敏
masked_prompt = masker.mask(raw_prompt)
print(f"脱敏后的 Prompt: {masked_prompt}")
# 输出: 请发送密码重置链接至 __MASKED_EMAIL_0__ 并测试 IP __MASKED_IP_ADDRESS_0__ 的连通性。

# 第二步：将安全数据通过 n1n.ai 发送至大模型
llm_response = send_secure_llm_request(masked_prompt, model_provider="anthropic")

# 第三步：将模型返回的占位符还原为原始敏感数据
final_output = masker.unmask(llm_response)
print(f"还原后的最终输出: {final_output}")