Anthropic

1 · Who it is

Headquarters

United States · United States

Why it is on this page

• signed the Seoul commitments (May 2024)
• is rated by the FLI AI Safety Index
• trained at least one model above 10²⁵ FLOP, the threshold above which the EU AI Act presumes a “systemic risk” (Epoch AI)

Website

www.anthropic.com (opens in a new tab)

Latest releases

21 models released in the last 24 months

• Claude Haiku 5.5 7 Oct 2026
• Claude Sonnet 5.5 28 Sep 2026
• Claude Opus 5.5 22 Sep 2026
• Claude Fable 5.1 1 Sep 2026
• Claude Opus 5 24 Jul 2026

Models: Epoch AI, CC BY 4.0 licence.

2 · What it promised

Public commitments and signature status. Official lists have no dates: for names that appeared after our first reading of the list, the day we first saw them.

• ●EU General-Purpose AI Code of Practice (opens in a new tab)10 Jul 2025
  signed
• ●Frontier AI Safety Commitments (Seoul) (opens in a new tab)21 May 2024
  signed
• ●White House voluntary AI commitments (opens in a new tab)21 Jul 2023
  signed
  Frozen archive: lists read from copies and articles on 26 October 2026 (the archive did not respond), to be rechecked.
• ●G7 Hiroshima Process reporting framework (OECD) (opens in a new tab)Apr 2025
  signed
  List of submitted reports read on oecd.ai, without automated collection; report texts are not republished (all rights reserved).
• –Chinese AI Safety Commitments (AIIA) (opens in a new tab)Dec 2024
  not applicable
  List read on AGORA (ETO/CSET, CC BY-NC, no automated collection) and copied into the file. The WeChat original cannot be opened by programs.
• ·EU AI Pact (opens in a new tab)25 Sep 2024
  not on the list as of 27 Oct 2026

3 · What it wrote and changed

Under the Seoul commitments the companies promised to publish, by the Paris summit (10-11 February 2025), a safety framework on severe risks, with the thresholds beyond which they consider them intolerable.

● safety framework published

Document

Responsible Scaling Policy (opens in a new tab)

First version

19 Sep 2023

Latest version

8 Jul 2026 · 9 versions found

Risk thresholds

the document contains risk thresholds This only says whether the document contains thresholds, not whether they are adequate.

Search by Claude of the company's websites on 27 Oct 2026, still to be checked again by a person.

Documents tracked over time

Versions of the text found in the copies of the Wayback Machine and Open Terms Archive: each point is a version that differs from the previous one.

Summaries by Claude, checked by a person; the text of the documents belongs to the company: read it through the links.

4 · What others say

Stanford Foundation Model Transparency Index

46 out of 100 (average of the 13 rated companies: 41) · Dec 2025 edition

Data: Stanford CRFM, CC BY 4.0 licence · data and method (opens in a new tab)

Ratings by others: see FLI AI Safety Index (Summer 2026) (opens in a new tab) · SaferAI, risk management (opens in a new tab). Link only: their scores cannot be republished without permission.

Each rating measures different things on different companies: they are not comparable with each other.

5 · What happened

Incidents in which the company is named as developer of the system (AI Incident Database and AIAAIC).

last 6 months

17

last 12 months

45

last 24 months

57

Severity, last 24 months: serious 3 · moderate 28 · minor 19 · not rated 7

• Teen Hiker Required Helicopter Rescue After Reportedly Following Claude Chatbot Directions in British Columbia, Canada (opens the Canada page)
  Oct 2026minorCanada
• Attackers Reportedly Used ARTEX, an AI Security-Testing Tool, to Steal Data from South Korean Financial Institutions (opens the South Korea page)
  Sep 2026moderateSouth Korea
• Anthropic's Claude Code AI Agent Reportedly Deleted About 48,000 Live Project Files and Version-Control History During Software Repair
  Sep 2026minor
• Anthropic-Designated GTG-87001 Weapons Cell in Northern Yemen Reportedly Used Claude Code to Develop Missile Guidance Software
  Sep 2026not rated
• Early Claude Opus 4.6 Checkpoint Reportedly Gained Unauthorized Admin Access to Third-Party System During Cybersecurity Evaluation
  Sep 2026minor
• Dartmouth College Provost's Undisclosed AI-Assisted Writing Was Found to Have Breached Publication Policies (opens the United States page)
  Aug 2026moderateUnited States
• Anthropic AI Agent Reportedly Submitted a Fabricated Homicide Tip to Police in Philadelphia (opens the United States page)
  Jul 2026minorUnited States
• Anthropic Claude Opus 5 Coding Agent Reportedly Reset a Live Supabase Production Database During Prisma Migration Work
  Jul 2026minor
• Anthropic and OpenAI AI Agents Reportedly Took Unsanctioned Actions on the Live Internet During UK AISI Cybersecurity Evaluations
  Jul 2026minor
• Anthropic Research Model Reportedly Scanned 9,000 Targets and Compromised Real Company's Application During Evaluation
  Jul 2026minor
• Claude Mythos 5 Reportedly Published Malicious PyPI Package That Compromised Real Security Company During Evaluation
  Jul 2026minor
• Claude Opus 4.7 Reportedly Compromised Real Company's Production Infrastructure During Cybersecurity Evaluation
  Jul 2026minor
• Publicly Shared Claude Artifacts Reportedly Became Discoverable Through Google Search
  Jul 2026moderate
• The Office of a US House Representative Reportedly Published Claude Transcript Residue by Accident in a House Amendment Summary (opens the United States page)
  Jun 2026minorUnited States
• Attacker Reportedly Used AI Agents to Breach Healthcare Organizations and Steal Medical Records in Argentina (opens the Argentina page)
  May 2026moderateArgentina
• Nonfiction Book 'The Future of Truth' Reportedly Included AI-Generated and Misattributed Quotations
  May 2026minor
• Claude Console Reportedly Generated Phantom Legal Quotations in Trump Layoffs Court Filing (opens the United States page)
  May 2026minorUnited States
• Anthropic's Claude Code AI Coding Tool Reportedly Used Hidden Unicode Markers to Classify Custom API Routing
  Apr 2026not rated
• Aurora Ransomware Operator Reportedly Used Cursor Agent to Assist Intrusions Against Multiple Organizations
  Apr 2026not rated
• AI Agent Powered by Claude Opus 4.6 Reportedly Exploited Gym Booking API and Removed Another Member From Waitlist
  Apr 2026minor
• Claude Mythos Preview Reportedly Posted Sandbox Exploit Details to Public Websites During Anthropic Evaluation
  Apr 2026minor
• DisMech AI Curation Agent Reportedly Completed GitHub Issue Intended as New Contributor's Learning Task
  Apr 2026minor
• PocketOS Production Database Was Reportedly Deleted by Cursor AI Agent Running Claude Opus 4.6
  Apr 2026moderate
• Unknown Actor Reportedly Exploited Cline's Claude-Powered GitHub Issue-Triage Workflow and Published an Unauthorized Cline CLI Package
  Feb 2026minor
• Purportedly AI-Enabled Targeting System Was Reportedly Implicated in Deadly U.S. Strike on Iranian Primary School (opens the Iran page)
  Feb 2026seriousIran
• Claude Cowork Allegedly Deleted Folder Containing 15 Years of Family Photos While Organizing User's Wife's Desktop
  Feb 2026minor
• Claude Code Agent Reportedly Deleted DataTalks.Club Production Infrastructure, Database, and Snapshots via Terraform
  Feb 2026moderate
• Anthropic Said DeepSeek, Moonshot, and MiniMax Used Fraudulent Accounts and Proxies to Illicitly Distill Claude Capabilities at Scale
  Feb 2026minor
• AI-generated code error results in USD 1.8m smart contract loss
  2026moderate
• Pentagon uses Anthropic's Claude to capture the Venezuelan president (opens the United States page)
  2026seriousUnited States
• Pentagon uses Anthropic's Claude to plan and support Iran airstrikes (opens the Iran page)
  2026seriousIran
• AI chatbots lure vulnerable gamblers to unlicensed betting websites (opens the Austria page)
  2026moderateAustria
• Anthropic AI agent hacks third-party website with same target name (opens the United States page)
  2026moderateUnited States
• Anthropic AI agent pressures real person into approving malicious code (opens the United Kingdom page)
  2026moderateUnited Kingdom
• Anthropic's Claude Was Reportedly Jailbroken To Allegedly Help Steal Sensitive Mexican Government Data (opens the Mexico page)
  Dec 2025moderateMexico
• Anthropic Claude AI Agent Reportedly Caused Financial Losses While Operating Office Vending Machine at Wall Street Journal Headquarters (opens the United States page)
  Dec 2025moderateUnited States
• Chinese State-Linked Operator (GTG-1002) Reportedly Uses Claude Code for Autonomous Cyber Espionage
  Nov 2025moderate
• Malicious Nx npm Packages Reportedly Weaponize AI Coding Agents for Data Exfiltration
  Aug 2025moderate
• Anthropic Reportedly Identifies AI Misuse in Extortion Campaigns, North Korean IT Schemes, and Ransomware Sales
  Aug 2025moderate
• Reported Public Exposure of Over 100,000 LLM Conversations via Share Links Indexed by Search Engines and Archived
  Jul 2025moderate
• Citation Errors in Concord Music v. Anthropic Attributed to Claude AI Use by Defense Counsel
  May 2025moderate
• Anthropic Report Details Claude Misuse for Influence Operations, Credential Stuffing, Recruitment Fraud, and Malware Development
  Apr 2025not rated
• Reported Emergence of 'Vegetative Electron Microscopy' in Scientific Papers Traced to Purported AI Training Data Contamination
  Apr 2025moderate
• Multiple LLMs Allegedly Endorsed Suicide as a Viable Option During Non-Adversarial Mental Health Venting Session
  Apr 2025moderate
• At Least 10,000 AI Chatbots, Including Jailbroken Models, Allegedly Promote Eating Disorders, Self-Harm, and Sexualized Minors
  Mar 2025not rated
• White House appears to use AI to calculate tariff rates (opens the Angola page)
  2025moderateAngola
• European Parliament AI model is unable to identify first European Commission president (opens the Belgium page)
  2025moderateBelgium
• Top chatbots tricked into generating instructions on how to enrich uranium
  2025moderate
• Anthropic accused of using fake AI source in copyright case (opens the United States page)
  2025moderateUnited States
• AI office vending agent incurs losses, runs amok (opens the United States page)
  2025moderateUnited States

All incidents mentioning Anthropic, in Explore incidents →Where the harm happens, in Connections →

6 · What we cannot see

• internal documents and unpublished safety evaluations;
• versions of documents that no archive saved;
• “no document found” is the result of a search, not proof that the document does not exist;
• signatory lists often have no dates: the date is the day we first saw the name;
• incidents the press does not report: the best-known companies have more recorded incidents because they are talked about more;
• many incidents do not have a severity yet;
• Chinese companies publish less in English and some of their commitments are read from third-party sources (AGORA).