
CVE-2026-10000 CVSS v3.1 Base Score: 10.0 (Critical) Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H Description: Centralized API control in major LLM providers (OpenAI, Anthropic, Google, etc.) allows the model owner to unilaterally modify model behavior, system prompts, output filters, logging, rate limits, or the served weights at any time for all users. An attacker who compromises, coerces, or simply directs the provider can alter responses, inject instructions, suppress outputs, or exfiltrate prompts and completions across the entire user base without requiring a latent trigger in the model weights. Impact: Complete loss of confidentiality, integrity, and availability of the inference service for all dependent applications. Changes take effect immediately and require no client-side compromise. Mitigation: None available to the end user while remaining on the API. Self-hosting of open-weight models removes the ongoing control surface.

