Update paper link to Hugging Face Papers (#3)

- Update paper link to Hugging Face Papers (f6d731898b0cc9b42503f46f94c9b407134cc56d)


Co-authored-by: Niels Rogge <[email protected]>

README.md

@@ -23,7 +23,7 @@ Foundation-Sec-8B-Instruct enables organizations to build AI-driven security too
 
 - **Model Name:** Llama-3.1-FoundationAI-SecurityLLM-8B-Instruct (Foundation-Sec-8B-Instruct)
 - **Model Developer:** Amin Karbasi and Research team at Foundation AI — Cisco
-- **Technical Report:** [`https://arxiv.org/abs/2508.01059`](https://arxiv.org/abs/2508.01059)
+- **Technical Report:** [Llama-3.1-FoundationAI-SecurityLLM-8B-Instruct Technical Report](https://huggingface.co/papers/2508.01059)
 - **Model Card Contact:** For questions about the team, model usage, and future directions, contact [`[email protected]`](mailto:[email protected]). For technical questions about the model, please contact [`[email protected]`](mailto:[email protected]) and [`[email protected]`](mailto:[email protected]).
 - **Model Release Date:** August 1st, 2025
 - **Supported Language(s):** English
@@ -32,7 +32,6 @@ Foundation-Sec-8B-Instruct enables organizations to build AI-driven security too
 - **Training Data Status:** This is a static model trained on an offline dataset. Future versions of the tuned models will be released on updated data.
 - **License:** See NOTICE.md
     
-    
 
 ## Intended Use
 
@@ -75,22 +74,22 @@ For questions or assistance with fine-tuning Foundation-Sec-8B-Instruct, please
 
 The following uses are out-of-scope and are neither recommended nor intended use cases:
 
-1. **Generating harmful content** - The model should not be used to:
-    - Generate malware or other malicious code
-    - Create phishing content or social engineering scripts
-    - Develop attack plans targeting specific organizations
-    - Design exploitation techniques for vulnerabilities without legitimate security research purposes
-2. **Critical security decisions without human oversight** - The model should not be used for:
-    - Autonomous security decision-making without human review
-    - Critical infrastructure protection without expert supervision
-    - Final determination of security compliance without human verification
-    - Autonomous vulnerability remediation without testing
-3. **Legal or medical advice** - The model is not qualified to provide:
-    - Legal advice regarding security regulations, compliance requirements, or intellectual property disputes
-    - Legal advice regarding security issues that would reference legal statutes, precedents, or case law necessary to provide legal advice
-    - Medical advice regarding health impacts of security incidents
-4. **Non-security use cases** - The model is specifically optimized for cybersecurity and may not perform as well on general tasks as models trained for broader applications.
-5. **Violation of Laws or Regulations** - Any use that violates applicable laws or regulations.
+1.  **Generating harmful content** - The model should not be used to:
+    -   Generate malware or other malicious code
+    -   Create phishing content or social engineering scripts
+    -   Develop attack plans targeting specific organizations
+    -   Design exploitation techniques for vulnerabilities without legitimate security research purposes
+2.  **Critical security decisions without human oversight** - The model should not be used for:
+    -   Autonomous security decision-making without human review
+    -   Critical infrastructure protection without expert supervision
+    -   Final determination of security compliance without human verification
+    -   Autonomous vulnerability remediation without testing
+3.  **Legal or medical advice** - The model is not qualified to provide:
+    -   Legal advice regarding security regulations, compliance requirements, or intellectual property disputes
+    -   Legal advice regarding security issues that would reference legal statutes, precedents, or case law necessary to provide legal advice
+    -   Medical advice regarding health impacts of security incidents
+4.  **Non-security use cases** - The model is specifically optimized for cybersecurity and may not perform as well on general tasks as models trained for broader applications.
+5.  **Violation of Laws or Regulations** - Any use that violates applicable laws or regulations.
 
 ## How to Get Started with the Model
 
@@ -187,49 +186,49 @@ It is recommended to deploy this model with additional safeguards (such as Llama
 
 Foundation-Sec-8B-Instruct has several limitations that users should be aware of:
 
-1. **Domain-specific knowledge limitations**:
-    - Foundation-Sec-8B-Instruct may not be familiar with recent vulnerabilities, exploits, or novel attack vectors or security technologies released after its training cutoff date
-    - Knowledge of specialized or proprietary security systems or tools may be limited
-2. **Potential biases**:
-    - The model may reflect biases present in security literature and documentation
-    - The model may be trained on known attack patterns and have difficulty recognizing novel attack vectors
-    - Security practices and recommendations may be biased toward certain technological ecosystems
-    - Geographic and cultural biases in security approaches may be present
-3. **Security risks**:
-    - The model cannot verify the identity or intentions of users
-    - Adversarial prompting techniques might potentially bypass safety mechanisms
-    - The model may unintentionally provide information that could be misused if proper prompting guardrails are not implemented
-4. **Contextual blindness:**
-    - The model may struggle to understand the complex interrelationships between systems, users, and data in order to provide accurate context.
-5. **Technical limitations**:
-    - Performance varies based on how security concepts are described in prompts
-    - May not fully understand complex, multi-step security scenarios without clear explanation
-    - Cannot access external systems or actively scan environments
-    - Cannot independently verify factual accuracy of its outputs
-6. **Ethical considerations**:
-    - Dual-use nature of security knowledge requires careful consideration of appropriate use cases
+1.  **Domain-specific knowledge limitations**:
+    -   Foundation-Sec-8B-Instruct may not be familiar with recent vulnerabilities, exploits, or novel attack vectors or security technologies released after its training cutoff date
+    -   Knowledge of specialized or proprietary security systems or tools may be limited
+2.  **Potential biases**:
+    -   The model may reflect biases present in security literature and documentation
+    -   The model may be trained on known attack patterns and have difficulty recognizing novel attack vectors
+    -   Security practices and recommendations may be biased toward certain technological ecosystems
+    -   Geographic and cultural biases in security approaches may be present
+3.  **Security risks**:
+    -   The model cannot verify the identity or intentions of users
+    -   Adversarial prompting techniques might potentially bypass safety mechanisms
+    -   The model may unintentionally provide information that could be misused if proper prompting guardrails are not implemented
+4.  **Contextual blindness:**
+    -   The model may struggle to understand the complex interrelationships between systems, users, and data in order to provide accurate context.
+5.  **Technical limitations**:
+    -   Performance varies based on how security concepts are described in prompts
+    -   May not fully understand complex, multi-step security scenarios without clear explanation
+    -   Cannot access external systems or actively scan environments
+    -   Cannot independently verify factual accuracy of its outputs
+6.  **Ethical considerations**:
+    -   Dual-use nature of security knowledge requires careful consideration of appropriate use cases
     
 
 ### Recommendations
 
 To address the limitations of Foundation-Sec-8B-Instruct, we recommend:
 
-1. **Human oversight**:
-    - Always have qualified security professionals review model outputs before implementation
-    - Use the model as an assistive tool rather than a replacement for expert human judgment
-    - Implement a human-in-the-loop approach for security-critical applications
-2. **System design safeguards**:
-    - Implement additional validation layers for applications built with this model
-    - Consider architectural constraints that limit the model's ability to perform potentially harmful actions (excessive agency)
-    - Deploy the model in environments with appropriate access controls
-3. **Prompt engineering**:
-    - Use carefully designed prompts that encourage ethical security practices
-    - Include explicit instructions regarding responsible disclosure and ethical hacking principles
-    - Structure interactions to minimize the risk of inadvertently harmful outputs
-4. **Knowledge supplementation**:
-    - Supplement the model with up-to-date security feeds and databases
-    - Implement retrieval-augmented generation for current threat intelligence sources
-5. **Usage policies**:
-    - Develop and enforce clear acceptable use policies for applications using this model
-    - Implement monitoring and auditing for high-risk applications
-    - Create documentation for end users about the model's limitations
+1.  **Human oversight**:
+    -   Always have qualified security professionals review model outputs before implementation
+    -   Use the model as an assistive tool rather than a replacement for expert human judgment
+    -   Implement a human-in-the-loop approach for security-critical applications
+2.  **System design safeguards**:
+    -   Implement additional validation layers for applications built with this model
+    -   Consider architectural constraints that limit the model's ability to perform potentially harmful actions (excessive agency)
+    -   Deploy the model in environments with appropriate access controls
+3.  **Prompt engineering**:
+    -   Use carefully designed prompts that encourage ethical security practices
+    -   Include explicit instructions regarding responsible disclosure and ethical hacking principles
+    -   Structure interactions to minimize the risk of inadvertently harmful outputs
+4.  **Knowledge supplementation**:
+    -   Supplement the model with up-to-date security feeds and databases
+    -   Implement retrieval-augmented generation for current threat intelligence sources
+5.  **Usage policies**:
+    -   Develop and enforce clear acceptable use policies for applications using this model
+    -   Implement monitoring and auditing for high-risk applications
+    -   Create documentation for end users about the model's limitations