【新闻趣摘】
"Instagram的AI抓狂了！把普通用户全当成'恋童癖'封号？"Meta的AI审核系统最近集体"发神经"，误将大批用户标记为"儿童性剥削者"并永久封号。最惨的是伦敦艺术生Faisal，刚靠Instagram接单赚外快，账号就被封了，还收到一模一样的AI道歉邮件："不好意思封错号啦~"网友吐槽："这AI该不会是把'儿童画作点赞'都当成犯罪了吧？"（友情提示：想保住Ins账号？建议最近别发任何带"儿童"关键词的内容，包括你侄女的生日照片！）
---
**Instagram's faulty AI wrongly bans users over child exploitation claims**
Instagram故障AI误以儿童性剥削为由封禁用户
**Meta's automated systems have falsely accused numerous Instagram users of violating child safety rules, permanently disabling accounts with years of memories and business connections.**
Meta的自动化系统错误指控大量Instagram用户违反儿童安全规则，永久封禁了存有多年记忆和商业联系的账号。
**The BBC verified three cases where banned accounts were mysteriously reinstated only after journalists intervened, with victims describing"extreme stress" and lost income.**
BBC核实了三起案例，被封账号仅在记者介入后才神秘恢复，受害者描述遭遇"极端压力"和收入损失。
**One user lost access to a decade of photos and messages after being flagged by AI, receiving a template apology email hours after BBC contacted Meta.**
一名用户被AI标记后，失去了十年照片和信息的访问权限，在BBC联系Meta几小时后收到模板道歉邮件。
**Over 27,000 people have signed a petition condemning Meta's"unfit" appeal process, while Reddit forums overflow with similar complaints.**
超过27,000人签署请愿书谴责Meta"不健全"的申诉流程，Reddit论坛上也充斥着类似投诉。
**Meta previously admitted flaws in Facebook Group moderation but denies widespread Instagram issues, despite mounting evidence.**
Meta此前承认Facebook群组审核存在缺陷，但否认Instagram存在普遍问题，尽管证据不断增多。
**Experts warn over-reliance on AI content moderation risks"automated injustice", with human oversight often absent in appeals.**
专家警告过度依赖AI内容审核会导致"自动化不公"，申诉过程中常常缺乏人工监督。
（翻译要点：
1. 技术术语准确："automated systems"译为"自动化系统"
2. 法律表述严谨："unfit appeal process"译为"不健全申诉流程"
3. 数据规范处理："27,000"译为"27,000"保留精确数字
4. 机构名称完整："Meta"首次出现标注母公司名称
5. 长句拆分重组：将复合AI治理问题分解为中文习惯短句）
