AI 生成诉讼潮：当 AI 让起诉变得太容易

"""
AI 法律文书生成系统（简化版 RAG 实现）
用于教学演示，不构成法律建议
"""
import json
from typing import List, Dict, Optional
from dataclasses import dataclass
from enum import Enum

class DocumentType(Enum):
    COMPLAINT = "起诉状"
    MOTION = "动议"
    ANSWER = "答辩状"
    BRIEF = "法律意见书"

@dataclass
class CaseFact:
    """案件事实"""
    plaintiff: str          # 原告
    defendant: str          # 被告
    jurisdiction: str       # 管辖法院
    claims: List[str]       # 诉讼请求
    facts: List[str]        # 事实描述
    damages: Optional[float] = None  # 索赔金额

@dataclass
class LegalReference:
    """法律引用"""
    citation: str           # 引用编号
    title: str              # 法律/判例名称
    relevance: str          # 与案件的关联性
    quoted_text: str        # 引用原文

class LegalKnowledgeBase:
    """法律知识库（模拟 RAG 检索）"""
    
    def __init__(self):
        self.statutes = []
        self.cases = []
    
    def add_statute(self, name: str, section: str, text: str):
        self.statutes.append({
            "name": name,
            "section": section,
            "text": text
        })
    
    def add_case(self, citation: str, name: str, holding: str):
        self.cases.append({
            "citation": citation,
            "name": name,
            "holding": holding
        })
    
    def search(self, keywords: List[str]) -> List[LegalReference]:
        """检索相关法律条文和判例（简化实现）"""
        results = []
        for statute in self.statutes:
            if any(kw in statute["text"].lower() for kw in keywords):
                results.append(LegalReference(
                    citation=f"{statute['name']} § {statute['section']}",
                    title=statute["name"],
                    relevance=f"涉及{'、'.join(keywords)}",
                    quoted_text=statute["text"]
                ))
        for case in self.cases:
            if any(kw in case["holding"].lower() for kw in keywords):
                results.append(LegalReference(
                    citation=case["citation"],
                    title=case["name"],
                    relevance=f"判例确立{'、'.join(keywords)}原则",
                    quoted_text=case["holding"]
                ))
        return results

class LegalDocumentGenerator:
    """法律文书生成器"""
    
    def __init__(self, knowledge_base: LegalKnowledgeBase):
        self.kb = knowledge_base
    
    def generate_complaint(self, case: CaseFact) -> str:
        """生成起诉状"""
        # 检索相关法律依据
        keywords = self._extract_keywords(case)
        references = self.kb.search(keywords)
        
        # 构建起诉状
        complaint = []
        complaint.append(f"{'=' * 60}")
        complaint.append(f"民事起诉状")
        complaint.append(f"{'=' * 60}")
        complaint.append(f"\n管辖法院：{case.jurisdiction}")
        complaint.append(f"原告：{case.plaintiff}")
        complaint.append(f"被告：{case.defendant}")
        complaint.append(f"\n{'─' * 40}")
        complaint.append(f"一、诉讼请求")
        for i, claim in enumerate(case.claims, 1):
            complaint.append(f"  {i}. {claim}")
        if case.damages:
            complaint.append(f"  索赔金额：{case.damages:,.2f} 美元")
        
        complaint.append(f"\n{'─' * 40}")
        complaint.append(f"二、事实陈述")
        for i, fact in enumerate(case.facts, 1):
            complaint.append(f"  {i}. {fact}")
        
        complaint.append(f"\n{'─' * 40}")
        complaint.append(f"三、法律依据")
        if references:
            for ref in references:
                complaint.append(f"\n  引用：{ref.citation}")
                complaint.append(f"  名称：{ref.title}")
                complaint.append(f"  关联：{ref.relevance}")
                complaint.append(f"  原文：「{ref.quoted_text[:100]}...」")
        else:
            complaint.append("  未找到直接相关的法律引用")
        
        complaint.append(f"\n{'─' * 40}")
        complaint.append(f"四、救济请求")
        complaint.append(f"  基于上述事实与法律依据，原告请求法院：")
        complaint.append(f"  1. 判令被告赔偿原告损失")
        complaint.append(f"  2. 判令被告承担诉讼费用")
        complaint.append(f"  3. 法院认为适当的的其他救济")
        complaint.append(f"\n{'=' * 60}")
        
        return "\n".join(complaint)
    
    def _extract_keywords(self, case: CaseFact) -> List[str]:
        """从案件事实中提取检索关键词"""
        keywords = []
        claim_text = " ".join(case.claims).lower()
        fact_text = " ".join(case.facts).lower()
        
        keyword_map = {
            "discrimination": ["discrimination", "歧视", "unequal treatment"],
            "employment": ["employment", "雇佣", "termination", "termination"],
            "contract": ["contract", "合同", "breach", "违约"],
            "negligence": ["negligence", "过失", "duty of care"],
            "copyright": ["copyright", "版权", "infringement"],
        }
        
        for category, terms in keyword_map.items():
            if any(term in claim_text or term in fact_text for term in terms):
                keywords.append(category)
        
        return keywords

# ===== 使用示例 =====
if __name__ == "__main__":
    # 初始化法律知识库
    kb = LegalKnowledgeBase()
    kb.add_statute("Civil Rights Act", "§ 1983", 
        "任何人在州法律授权下剥夺公民宪法权利，应承担法律责任")
    kb.add_statute("Clean Air Act", "§ 7413",
        "违反空气质量标准的排放行为，环保署可处以民事处罚")
    kb.add_case("582 U.S. ___ (2024)", "Environmental Justice Coalition v. EPA",
        "环境监管中的种族差异构成违宪歧视")
    
    # 构建案件事实
    case = CaseFact(
        plaintiff="原告：社区代表 John Doe",
        defendant="被告：某科技公司",
        jurisdiction="美国联邦地区法院 - 马萨诸塞州",
        claims=["违反环境法规", "社区健康损害", "要求停止排放并赔偿"],
        facts=[
            "被告在居民区运营大型数据中心",
            "24 台燃气轮机 24 小时运转，排放污染物",
            "社区居民呼吸道疾病发病率显著上升",
            "被告未取得有效空气排放许可"
        ],
        damages=5000000.0
    )
    
    # 生成起诉状
    generator = LegalDocumentGenerator(kb)
    complaint = generator.generate_complaint(case)
    print(complaint)

"""
AI 生成法律文书检测系统
帮助法院识别和审查 AI 生成的法律文件
"""
from typing import Dict, List, Tuple
from dataclasses import dataclass
import re

@dataclass
class DetectionResult:
    """检测结果"""
    is_ai_generated: bool       # 是否 AI 生成
    confidence: float           # 置信度 0-1
    indicators: List[str]      # 检测指标
    hallucination_risk: str    # 幻觉风险等级
    
class AIDocumentDetector:
    """AI 生成法律文件检测器"""
    
    def __init__(self):
        self.ai_patterns = [
            # 过度正式的模板语言
            r"pursuant to the provisions of",
            r"in accordance with applicable laws and regulations",
            r"the undersigned hereby respectfully requests",
            # 重复的论证结构
            r"firstly.*secondly.*thirdly.*fourthly",
            # 过度使用法律术语
            r"inter alia|prima facie|ipso facto|res ipsa loquitur",
        ]
        self.citation_db = self._build_citation_database()
    
    def analyze_document(self, text: str) -> DetectionResult:
        """分析法律文件是否由 AI 生成"""
        indicators = []
        ai_score = 0.0
        
        # 检测指标 1: 语言模式
        language_score = self._check_language_patterns(text)
        if language_score > 0.6:
            indicators.append("语言模式异常正式/模板化")
            ai_score += language_score * 0.3
        
        # 检测指标 2: 结构一致性
        structure_score = self._check_structure_consistency(text)
        if structure_score > 0.7:
            indicators.append("文档结构高度标准化")
            ai_score += structure_score * 0.2
        
        # 检测指标 3: 法律引用验证
        citation_issues = self._verify_citations(text)
        if citation_issues:
            indicators.extend(citation_issues)
            ai_score += 0.3
        
        # 检测指标 4: 论证逻辑
        logic_score = self._check_logical_coherence(text)
        if logic_score < 0.5:
            indicators.append("论证逻辑存在断裂")
            ai_score += 0.2
        
        # 综合判断
        is_ai = ai_score > 0.5
        confidence = min(ai_score, 1.0)
        
        # 幻觉风险评估
        if citation_issues:
            hallucination_risk = "高"
        elif ai_score > 0.7:
            hallucination_risk = "中"
        else:
            hallucination_risk = "低"
        
        return DetectionResult(
            is_ai_generated=is_ai,
            confidence=confidence,
            indicators=indicators,
            hallucination_risk=hallucination_risk
        )
    
    def _check_language_patterns(self, text: str) -> float:
        """检查语言模式"""
        matches = sum(1 for p in self.ai_patterns if re.search(p, text, re.IGNORECASE))
        return min(matches / len(self.ai_patterns), 1.0)
    
    def _check_structure_consistency(self, text: str) -> float:
        """检查结构一致性（简化实现）"""
        # 检查段落长度一致性
        paragraphs = text.split('\n\n')
        if len(paragraphs) < 3:
            return 0.0
        lengths = [len(p) for p in paragraphs if len(p.strip()) > 0]
        if not lengths:
            return 0.0
        avg = sum(lengths) / len(lengths)
        variance = sum((l - avg) ** 2 for l in lengths) / len(lengths)
        # 方差越小，结构越一致（越可能是 AI 生成）
        return max(0, 1 - variance / (avg * 100))
    
    def _verify_citations(self, text: str) -> List[str]:
        """验证法律引用（简化实现）"""
        issues = []
        # 检查引用格式
        citation_pattern = r'\d+\s+U\.S\.\s+\d+'
        citations = re.findall(citation_pattern, text)
        for cite in citations:
            if cite not in self.citation_db:
                issues.append(f"⚠️ 未找到引用: {cite}（可能为虚构引用）")
        return issues
    
    def _check_logical_coherence(self, text: str) -> float:
        """检查论证逻辑连贯性（简化实现）"""
        # 检查论点与论据的对应关系
        transition_words = [
            "therefore", "thus", "consequently", "hence",
            "however", "nevertheless", "moreover", "furthermore"
        ]
        transitions = sum(1 for w in transition_words if w in text.lower())
        # 过渡词过多可能表示 AI 生成的"伪论证"
        if transitions > 20:
            return 0.3
        return 0.8
    
    def _build_citation_database(self) -> set:
        """构建引用数据库（简化版本）"""
        return {
            "582 U.S. ___ (2024)",
            "509 U.S. 579 (1993)",
            "437 U.S. 678 (1978)",
        }

# ===== 使用示例 =====
if __name__ == "__main__":
    detector = AIDocumentDetector()
    
    sample_doc = """
    民事起诉状
    pursuant to the provisions of the Federal Rules of Civil Procedure,
    the undersigned hereby respectfully requests that this Court
    enter judgment in favor of the plaintiff.
    
    Firstly, the defendant violated the Clean Air Act.
    Secondly, the defendant caused environmental harm.
    Thirdly, the defendant failed to obtain permits.
    Fourthly, the community suffered health damages.
    
    Citation: 582 U.S. ___ (2024) established environmental justice.
    Citation: 999 U.S. 123 (2025) supports this claim.
    """
    
    result = detector.analyze_document(sample_doc)
    print(f"AI 生成: {result.is_ai_generated}")
    print(f"置信度: {result.confidence:.2f}")
    print(f"指标: {result.indicators}")
    print(f"幻觉风险: {result.hallucination_risk}")