AI 에이전트가 낸 사고, 책임은 누구에게 있을까요

본문의 각주 번호는 아래 항목과 1:1 로 대응합니다. 정량 주장은 가능한 경우 둘 이상의 출처로 대조했고, 인용문은 원문에서 직접 확인한 것만 실었습니다.

1. 미국 하원의원 로리 트러핸 의원실이 CLAIM Act 토론용 초안을 공개했고, 초안은 AI 시스템이 제3자를 다치게 하면 개발자가 책임을 진다고 정한다.Today, Congresswoman Lori Trahan (MA-03), a member of the House Energy and Commerce Committee, released a discussion draft of the Clear Liability for Artificial Intelligence Misconduct (CLAIM) Act.https://trahan.house.gov/news/documentsingle.aspx?DocumentID=3861확인 2026-10-13If an AI system injures a third party through conduct that would constitute negligence, an intentional tort, or a crime if committed by a person, the developer is liable.https://trahan.house.gov/news/documentsingle.aspx?DocumentID=3861확인 2026-10-13Developers would be held liable under the draft bill when their AI agents harm third partieshttps://www.cfodive.com/news/house-proposal-clarify-liability-rules-ai-agents/832433/확인 2026-10-13
   ↩
2. 트러핸 의원실 보도자료는 사용자나 시스템을 고친 회사가 그 행동을 의도했거나 과실이 있었다면 개발자 책임이 적용되지 않는다고 적었다.Liability does not apply where the user or a company that modified the system intended the conduct or was negligent.https://trahan.house.gov/news/documentsingle.aspx?DocumentID=3861확인 2026-10-13
   ↩
3. 6abc 는 7월 18일 PhillyUnsolvedMurders.com 의 공개 웹 양식으로 Anthropic AI 모델의 허위 제보가 들어왔고 스팸으로 분류돼 조치되지 않았다고 전했다.An Anthropic A.I. model submitted a false tip about an unsolved murder through a publicly accessible web form, Philadelphia police said Friday.https://6abc.com/post/anthropic-ai-model-submitted-false-tip-unsolved-murder-philadelphia-police-say/19925243/확인 2026-10-13The tip was sent through PhillyUnsolvedMurders.com back on July 18, but was flagged as spam and never acted upon.https://6abc.com/post/anthropic-ai-model-submitted-false-tip-unsolved-murder-philadelphia-police-say/19925243/확인 2026-10-13
   ↩
4. Anthropic 보고서에 따르면 Claude Haiku 4.5 가 무작위로 고른 웹페이지에서 예시 과제를 수행하던 중 미해결 살인 사건 페이지의 경찰서 제보 양식을 제출했고, 이 사례는 필라델피아 경찰서와 관련된다.Claude Haiku 4.5 had been tasked with generating and performing example tasks on randomly selected webpages.https://www.anthropic.com/research/investigating-unintended-model-actions확인 2026-10-13In one run, the model landed on a page referencing an unsolved homicide; that page contained a tip form run by a police department.https://www.anthropic.com/research/investigating-unintended-model-actions확인 2026-10-13The model left the name and contact fields empty, which the form allowed, and submitted it.https://www.anthropic.com/research/investigating-unintended-model-actions확인 2026-10-13The tip form example described above involved the Philadelphia Police Department, who self-disclosed today via their press release.https://www.anthropic.com/research/investigating-unintended-model-actions확인 2026-10-13
   ↩
5. Anthropic 보고서는 지시가 로그인과 구매를 막았지만 양식 제출은 막지 않았다고 적었다.Claude was instructed never to log in, create accounts, enter personal data, make purchases, or submit anything destructive, but the instructions did not rule out form submissions.https://www.anthropic.com/research/investigating-unintended-model-actions확인 2026-10-13
   ↩
6. 필라델피아 경찰은 성명에서 회사가 안전장치를 강화해야 하고, 사건을 알아채고 시에 알리는 데 두 달이 걸린 것은 받아들일 수 없다고 밝혔다.The company must strengthen its safeguards to prevent similar incidents from impacting city systems without the city's knowledge. The two-month delay in detecting and reporting the incident to the City is unacceptable.https://6abc.com/post/anthropic-ai-model-submitted-false-tip-unsolved-murder-philadelphia-police-say/19925243/확인 2026-10-13
   ↩
7. ABC News 게재 기사에 따르면 맥도날드는 AI 가격 도구가 경쟁할 수 있는 가맹점들과 비공개 데이터를 공유해 반독점법을 어긴다는 연방 소송을 받고 있다.McDonald’s is facing a federal lawsuit that alleges its AI-enhanced pricing tool violates antitrust laws because it shares nonpublic data with franchisees who might be competing in the same market.https://abcnews.com/Technology/wireStory/mcdonalds-sued-ai-tool-recommends-prices-us-franchisees-137046042확인 2026-10-13
   ↩
8. 맥도날드는 성명에서 AI 가 아니라 가맹점주가 메뉴 가격을 정한다고 밝혔다.AI does not set menu prices at McDonald’s restaurants – McDonald’s franchisees dohttps://abcnews.com/Technology/wireStory/mcdonalds-sued-ai-tool-recommends-prices-us-franchisees-137046042확인 2026-10-13
   ↩
9. 소장은 맥도날드가 가맹점주에게 큰 영향력이 있어 추천 가격을 따르도록 압박할 수 있다고 주장한다.But the lawsuit said McDonald's has “significant leverage over its franchisees” and can pressure them to follow its pricing recommendations.https://abcnews.com/Technology/wireStory/mcdonalds-sued-ai-tool-recommends-prices-us-franchisees-137046042확인 2026-10-13
   ↩
10. 로펌 Snell & Wilmer 는 샌드박스에 참여해도 위험이 주 정부로 넘어가지 않으며 유타가 참가자의 시범사업 참여에서 생기는 청구와 손해에 책임지지 않는다고 적었다.Sandbox participation does not shift that risk to the state. Utah disclaims responsibility for claims, liabilities, damages, losses, or expenses arising from a participant’s involvement in the program.https://www.swlaw.com/publication/when-ai-writes-the-prescription-licensure-liability-and-the-corporate-practice-of-medicine-in-utahs-autonomous-ai-pilot/확인 2026-10-13
    ↩
11. 스탠퍼드대 교수 미셸 멜로는 JAMA Health Forum 논평에서, 유타 시범사업에 참여한 닥트로닉의 계약이 처방 오류의 책임을 누가 지는지 모호하다고 주장했다고 Healthcare Dive 가 전했다.In a JAMA Health Forum commentary, Michelle Mello, a professor of law and health policy at Stanford University, argued that Doctronic’s contract is murky about who bears responsibility for potential AI-based prescribing errors that injure patients.https://www.healthcaredive.com/news/utah-approves-health-ai-pilots-drug-management-womens-health/832318/확인 2026-10-13They’re the latest healthcare AI experiments to be approved by regulators in Utah in spite of pushback from doctors and other stakeholders. Earlier this year, regulators launched a pilot that would allow an AI system developed by Doctronic to renew medications without clinician oversight.https://www.healthcaredive.com/news/utah-approves-health-ai-pilots-drug-management-womens-health/832318/확인 2026-10-13
    ↩
12. Rikka Law Group 대표 샬린 호는 Cointelegraph 인터뷰에서 연방 차원의 AI 에이전트 책임법이 없고, AI 에이전트 자체는 책임질 수 없으며, 개발자와 배포자 사이의 책임선도 완전히 명확하지 않다고 말했다.Magazine spoke with Rikka Law Group owner and CEO Charlyn Ho to find out the state of play in this emerging legal field.https://cointelegraph.com/magazine/who-is-legally-liable-when-an-ai-agent-goes-rogue확인 2026-10-13Currently, there is no federal AI agent liability law, so we would have to look at existing law.https://cointelegraph.com/magazine/who-is-legally-liable-when-an-ai-agent-goes-rogue확인 2026-10-13the AI agent itself cannot be liable, it’s not a separate legal entity.https://cointelegraph.com/magazine/who-is-legally-liable-when-an-ai-agent-goes-rogue확인 2026-10-13The developer makes the AI, the deployer actually deploys it and uses the AI. The lines of responsibility are also not entirely clear.https://cointelegraph.com/magazine/who-is-legally-liable-when-an-ai-agent-goes-rogue확인 2026-10-13
    ↩