Securing Ai Supply Chains Coheres Commitment To Model Signing

North Mini Code. Cohere's first model for developers.

Learn more

Oct 30, 2025

2 minutes read

Securing AI supply chains: Cohere’s commitment to model signing

Cohere has implemented model signing for all Cohere Command models hosted on Hugging Face to improve integrity and authenticity efforts.

!Blog Post Featured Image

In an era where AI models underpin critical enterprise decisions, the integrity and authenticity of these models have become paramount. Cohere is proud to announce a significant advancement in AI security and transparency: we’ve implemented model signing for all Cohere Command models hosted on Hugging Face, ensuring verifiable authenticity and integrity for our models on the platform. This initiative aligns with our leadership and involvement in the Coalition for Secure AI (CoSAI) and reinforces our commitment to securing the AI supply chain.

##### The challenge: Trust in AI deployment

The rapid integration of AI into enterprise environments has introduced novel supply chain risks. As organizations increasingly rely on pre-trained models, fine-tuned variants, and third-party AI services, they face a critical vulnerability: the AI model supply chain has become a prime attack vector. Unlike traditional software vulnerabilities, compromised AI models can:

• Propagate hidden biases that lead to discriminatory decisions in hiring, lending, or customer service
• Create backdoors that activate under specific conditions, bypassing security controls
• Leak training data or intellectual property through model inversion attacks

These failures and attacks can often manifest subtly. Enterprises need a robust mechanism to verify that the models they deploy are authentic, unaltered, and sourced from trusted producers.

##### Model signing: A foundation for trust

Model signing employs cryptographic techniques to create a tamper-proof digital signature for AI artifacts. This signature verifies three critical aspects:

###### Integrity

Ensures the model has not been modified since creation.

###### Provenance

Provides a traceable record of the model’s development lifecycle and dependencies.

###### Properties

Validates claims about performance, compliance, and other characteristics.

By adopting model signing, organizations can automate security policies, streamline regulatory compliance, and accelerate deployment while maintaining trust.

##### Cohere’s leadership in secure AI

As part of our commitment to secure AI, Cohere is publishing signing artifacts for all Command models. Users can now verify the authenticity and integrity of our models that are deployed directly on Hugging Face. This initiative sets a new standard for transparency and empowers developers and enterprises to build AI systems with confidence.

Here is how to verify a model signature using the model-signing library:

1. Install the model-signing package:
pip install model-signing

2. Install huggingface_hub and download the model:
pip install huggingface_hub
huggingface-cli download CohereLabs/c4ai-command-r7b-arabic-02-2025 --revision main --local-dir ./model

3. Verify the signature:
model_signing verify ./model \
--signature ./model/signatures/c4ai-command-r7b-arabic-02-2025.sig
--identity "https://github.com/cohere-ai/model-signing/.github/workflows/sign-model.yml@refs/heads/main" \
--identity_provider "https://token.actions.githubusercontent.com" \
--ignore_unsigned_files

Cohere’s efforts build on the groundbreaking work of CoSAI, which is establishing crucial standards for model signing—a vital step towards securing the AI supply chain. Alongside  other industry leaders working with CoSAI, we're implementing these practices to ensure model integrity and trustworthy AI ecosystems.

##### Download signed models now

Explore our signed Command and Cohere Labs Aya Vision models on Hugging Face. Navigate to the signature directory for each model, where the directory will contain the model signature (e.g. command-a-reasoning-08-2025.sig) and a verification-instructions.txt that has a step-by-step guide on how to run the verification locally. These types of efforts help build a safer AI future.

##### The path forward

As AI becomes embedded in critical infrastructure, model signing will be as essential as SSL certificates are for web security today. Cohere is committed to advancing this paradigm shift, ensuring AI supply chains remain secure, transparent, and trustworthy. To learn more about Cohere, schedule a demo with our team.