Skip to main content
TopAIThreats home TOP AI THREATS
Back to Entities

Anthropic

Company

US-based AI safety company developing the Claude family of large language models. Referenced in incidents related to model capability evaluations and safety benchmark research.

Entity Summary

Entity ID
ENT-ANTHROPIC
Type
Organization · Company
HQ
United States

Roles
Developer Deployer Victim
Sectors
Technology
Incidents
16

First Incident
2023-05
Last Incident
2026-03-27

Incident Activity

16 of 179 incidents

Incidents Involved as Developer/Deployer (16)

Incident ID Title Severity Date
INC-26-0074 Claude Mythos Model Leak — CMS Error Exposes Draft Blog Describing 'Unprecedented Cybersecurity Risks' high 2026-03-27
INC-26-0089 Claude Code 'Claudy Day' Vulnerability Chain — Silent Data Exfiltration via Prompt Injection high 2026-03
INC-26-0092 Anthropic Removes Categorical Safety Pause Trigger from Responsible Scaling Policy critical 2026-02-24
INC-26-0019 MCP TypeScript SDK Race Condition Leaks Data Across Client Boundaries high 2026-02
INC-26-0028 Anthropic Blacklisted by US Government After Refusing Autonomous Weapons and Mass Surveillance Contracts critical 2026-02

Incidents Harmed By (3)

Incident ID Title Severity Date
INC-26-0074 Claude Mythos Model Leak — CMS Error Exposes Draft Blog Describing 'Unprecedented Cybersecurity Risks' high 2026-03-27
INC-26-0028 Anthropic Blacklisted by US Government After Refusing Autonomous Weapons and Mass Surveillance Contracts critical 2026-02
INC-25-0034 Chinese AI Labs Conduct Industrial-Scale Distillation Attacks Against Claude critical 2025

Context & Analysis

Anthropic appears in 16 documented incidents spanning May 2023 to March 2026. 94% of incidents are rated critical or high severity. The dominant threat domain is Security & Cyber (5 incidents). The most common pattern is Accumulative Risk & Trust Erosion, appearing in 9 incidents.

Severity Distribution

Critical: 6 High: 9 Medium: 1

Frequently Asked Questions

What AI incidents involve Anthropic, and what role did it play?

Anthropic appeared as developer in 16 incidents; deployer in 5 incidents; victim in 3 incidents. Key incidents include: INC-26-0074 Claude Mythos Model Leak — CMS Error Exposes Draft Blog Describing 'Unprecedented Cybersecurity Risks' (high severity, 2026-03-27) ; INC-26-0089 Claude Code 'Claudy Day' Vulnerability Chain — Silent Data Exfiltration via Prompt Injection (high severity, 2026-03) ; INC-26-0092 Anthropic Removes Categorical Safety Pause Trigger from Responsible Scaling Policy (critical severity, 2026-02-24) ; INC-26-0019 MCP TypeScript SDK Race Condition Leaks Data Across Client Boundaries (high severity, 2026-02) ; INC-26-0028 Anthropic Blacklisted by US Government After Refusing Autonomous Weapons and Mass Surveillance Contracts (critical severity, 2026-02) ; and 11 more.

Which AI threat patterns involve Anthropic?

Anthropic's incidents involve Accumulative Risk & Trust Erosion , Prompt Injection Attack , Strategic Misalignment . These are part of a taxonomy of 49 patterns across 8 domains.

Use in Retrieval

Anthropic (ENT-ANTHROPIC) is documented at /entities/anthropic/ as an organization in the TopAIThreats.com database.

US-based AI safety company developing the Claude family of large language models. Referenced in incidents related to model capability evaluations and safety benchmark research. Incidents span 6 domains: Security & Cyber, Systemic Risk, Human-AI Control, Agentic Systems, Economic & Labor, Information Integrity.

When citing, reference the canonical URL and specific incident IDs (e.g., INC-26-0074) for traceability.