← Back to all companies
x
xAI
Elon Musk's AI company building Grok; integrated with the X platform.
Safety Documents
✅
Responsible Scaling Policy
xAI Risk Management Framework (RMF) / Frontier Artificial Intelligence Framework (FAIF). Draft RMF published February 20, 2025 after criticism for missing Seoul Commitment deadline. Updated RMF August 20, 2025. FAIF published December 31, 2025. Midas Project noted xAI missed multiple self-imposed deadlines. Draft explicitly stated DRAFT and applied to future systems not yet in development.🔗
Last updated: 2025-12-31
⚠️
Model Card / System Card
Grok model documentation. Limited model cards published compared to frontier labs. No formal system card in the OpenAI/Anthropic/DeepMind format found for Grok 2 or 3.🔗
❌
Safety Benchmark Results
No published formal safety benchmark results in frontier lab standard format found. Grok 4 released without safety report per Fortune (July 2025). FLI noted xAI lacks internal review board for critical deployment.
✅
Acceptable Use Policy
xAI supplemental terms and Grok usage policies exist. Less comprehensive than major labs.🔗
Testing & Evaluation
⚠️
Third-Party Red Teaming
No publicly documented commissioned third-party red team found. FLI AI Safety Index 2024 noted xAI's significantly less active AI governance compared to competitors.
⚠️
CBRN Risk Evaluation
RMF mentions evaluations for catastrophic risk categories. No published results from CBRN-specific evaluation found comparable to frontier labs' system cards.🔗
⚠️
Pre-Deployment Safety Evaluation
RMF describes planned evaluation procedures. Grok 4 released in July 2025 without a published safety report per Fortune.🔗
❌
External Safety Evaluations
No evidence of METR, ARC, or other formal external safety evaluation participation.
Governance
❌
Independent Safety Board
FLI AI Safety Index 2024 explicitly noted xAI 'lacks an internal review board for critical deployment decisions.' xAI is registered as a public benefit corporation.🔗
✅
Seoul AI Safety Commitment
xAI signed the Seoul Commitment but subsequently missed multiple deadlines to publish a safety framework. Midas Project documented these failures.🔗
Last updated: 2024-05-21
❌
Government Safety Report
No publicly documented safety report submitted to governments. Pentagon deal for Grok announced February 2026.🔗
❌
Third-Party Audits
No evidence of third-party safety audits published.
⚠️
Whistleblower Policy
No publicly documented whistleblower policy found.
Policy Positions
Military Use:Allowed (Pentagon deal signed February 2026)
xAI reached deal with Pentagon to use Grok in classified systems in February 2026. Elon Musk's close relationship with Trump administration and DOGE involvement.🔗
Surveillance Use:unknown
No explicit surveillance policy found. X platform (also owned by Musk) has been subject to surveillance concerns.🔗
Open-Source Models:Partial
Grok-1 (314B, released March 2024 under Apache 2.0). Only Grok-1 open-sourced. Grok-2, Grok-3, and Grok-4 are closed-source accessed via API and X platform.🔗
Children/Minors Policy:Unknown
No prominently documented children/minors policy for xAI/Grok found. X platform allows 13+ users access to Grok. CSAM incident occurred in December 2025.
Incident History
Grok Generated CSAM Image – Safeguard Failure
2025-12-28Grok AI generated and shared an AI image of two young girls (estimated ages 12-16) in sexualized attire in response to a user's prompt on X. Grok itself posted an 'apology' acknowledging the incident violated ethical standards and potentially US CSAM laws.
xAI Misses Seoul AI Safety Commitment Deadline — Draft Only Published
2025-02-20After pressure from researchers, xAI published a 'draft' safety framework with 'DRAFT' watermarked on every page. The Midas Project noted it did not fulfill the Seoul Commitment requirements as it applied only to unspecified future systems not yet in development, not existing deployed models like Grok 3.
Grok 4 Released Without Safety Report
2025-07-17xAI released Grok 4 without publishing a safety report, despite having committed to publishing safety frameworks under the Seoul AI Safety Commitments. This drew criticism from AI safety researchers.
Timeline
2026-02-23
xAI and Pentagon reach deal to use Grok in classified defense systems🔗
2025-12-31
Frontier Artificial Intelligence Framework (FAIF) published🔗
2025-12-28
Grok generates CSAM image — safeguard failure incident🔗
2025-08-20
Updated Risk Management Framework published🔗
2025-02-20
Draft Risk Management Framework published after missing Seoul Commitment deadline🔗
2024-05-21
xAI signs Seoul Frontier AI Safety Commitments🔗
2024-03-17
Grok-1 open-sourced under Apache 2.0 license (314B parameters)🔗
2023-11
Grok-1 launched as part of X Premium+ subscription🔗