Privacy Policy

Usage Policy

EULA

How We're Different

Our Platform

Readiness

Content

Engagement

Security

Life Sciences

Financial Services

Telecommunications

Manufacturing

Technology

Retail

Government

Energy

Sales Teams

Marketing Teams

Channel Partners

About us

Newsroom

Careers

Investors

Contact

Pricing

Partners

Bigtincan Support

Bigtincan Help Center

Get a demo

Log in

Speak to our team

Why Bigtincan?

Welcome to a new era of sales enablement.

Take the buying experience to new frontiers.

Products

Prepare and upskill your customer-facing teams.

Access, personalize, and share content with buyers.

Build meaningful engagement with powerful tools.

Custom tailor our platform to fit your organization's needs.

Solutions

Stay focused on sales activities and win more.

Bridge the gap between marketing and sales.

Boost revenue with your channel partners. 

Resources

Insights for high-performing sales teams.

Meet those transforming their buyers’ experiences.

Learn how we leverage our partner network.

Integrate Bigtincan with your current tech stack

Praesent commodo cursus magna, vel scelerisque nisl consectetur et. Fusce dapibus, tellus ac cursus commodo

<p>We understand that our customers have important questions regarding the security, fairness, and privacy of our AI-driven solutions. To address these, we’ve compiled a list of common inquiries about our AI practices and third-party integrations.</p>
<p>Here are some clear, detailed answers to ensure you are informed about how we approach AI, safeguard your data, and maintain ethical standards.</p>


Download the PDF version

<p>The third-party LLM providers, such as OpenAI, perform their own security assessments, which typically include vulnerability testing as part of their development and deployment processes. The SaaS company relies on the security assurances and certifications provided by these third-party providers. Details on specific vulnerability tests would be best obtained directly from the third-party provider.</p>


<p>The third-party LLM providers actively research and implement techniques to mitigate bias in their models. However, it is important to note that while efforts are made to reduce bias, it is not possible to eliminate all biases entirely. The third-party providers typically release documentation and research papers that describe their approach to bias mitigation.</p>


<p>The third-party LLM providers use a combination of data augmentation, diverse training datasets, and statistical techniques to test and mitigate biases that could arise from data limitations. These providers often publish their methodologies and research findings in technical documentation or academic papers, where they describe how they address issues like missing data, observational intensity, and demographic representation.</p>


<p>The bias and fairness testing methodologies employed by third-party LLM providers generally include:</p>
<ul>
<li>Diverse Training Data: Ensuring the training data includes a wide range of demographic groups.</li>
<li>Bias Detection Tools: Using automated tools to detect and measure bias during the model development process.</li>
<li>Iterative Model Training: Continuously refining models to reduce identified biases.</li>
</ul>
<p>These methodologies are typically outlined in the third-party provider’s public documentation.</p>


<div class="page" title="Page 2">
<div class="section">
<div class="layoutArea">
<div class="column">
<p>Third-party LLM providers often identify bias detection and fairness metrics early in the model design phase. These metrics guide the development and testing process to ensure the models perform fairly across different subgroups. The specifics of these metrics are generally included in the provider’s technical documentation or research publications.</p>
</div>
</div>
</div>
</div>


<div class="page" title="Page 2">
<div class="section">
<div class="layoutArea">
<div class="column">
<p>Third-party LLM providers typically use pre- identified metrics to test and measure model performance across various subgroups. This ongoing evaluation helps ensure that the models maintain fairness and reduce bias over time. Details on these metrics and their application are usually found in the provider’s documentation.</p>
</div>
</div>
</div>
</div>


<div class="page" title="Page 3">
<div class="section">
<div class="layoutArea">
<div class="column">
<p>Third-party LLM providers evaluate their models for fairness by analyzing the outputs for indications of unfair or harmful stereotypes. They use a combination of automated tools and human evaluations to identify and mitigate these issues. The results of these assessments are often summarized in the provider’s documentation or research papers.</p>
</div>
</div>
</div>
</div>


<p>TLS (Transport Layer Security) implementation details, including scan results, would typically be managed by the SaaS company if they handle the integration and data transport layers. For third-party LLM providers, TLS is generally part of their overall security infrastructure, and specific scan results would need to be obtained directly from them.</p>


<p>The SaaS company would rely on the third- party LLM providers’ assurances regarding their vulnerability management processes. Specific vulnerability scan results for the LLMs would be proprietary to the third-party provider and are not typically disclosed publicly.</p>


<p>The third-party LLM providers employ extensive validation processes to confirm the accuracy of their model outputs. These processes often include benchmark testing against industry standards, cross-validation techniques, and continuous refinement based on real-world feedback. The details of these processes are usually outlined in the provider’s technical documentation.</p>


<p>Third-party LLM providers continuously monitor and improve the accuracy of their models through a combination of real-time performance tracking, periodic re-training with new data, and user feedback integration. They regularly update their models to adapt to new data and use cases. This process is typically described in the provider’s documentation.</p>


<p>Yes, many third-party LLM providers have undergone independent statistical validation for the accuracy of their outputs. These validations are often conducted by external experts or through peer-reviewed research to ensure that the models meet industry standards for accuracy. The results of these validations are generally published in research papers or made available through the provider’s documentation.</p>
<p>These responses focus on factual information related to the third-party LLM providers, without making any assumptions about the SaaS company’s own practices beyond relying on the third-party provider’s assurances.</p>


<p>Bigtincan does not send customer data to third parties to train their models. When Bigtincan leverages third-party models, we ensure that available privacy and safety settings are configured to limit the disclosure, storage, and use of your data for model training purposes.</p>


<p>Bigtincan’s AI features can integrate with customer cloud infrastructure (as determined at discovery). For third-party managed models, including GPT models, we leverage managed cloud infrastructure like Microsoft Azure, Amazon Bedrock, and the like.</p>


<p>No. However, users feedback provide on the accuracy or effectiveness of an Al-based feature sets may be used to improve our AI offerings and/ or to inform future product development.</p>


<p>Yes, we prioritize transparency and the choice of Al in our featureset. Our AI Iconography or phrases like “powered by Genie” indicates our generative Al features in the interface, so users are informed when they engage with generative Al technology.</p>


<p>Bigtincan’s generative Al features are opt-in by default.</p>


<p>Our privacy standards are disclosed on our SecureGLP white paper. Bigtincan generative Al features meet Bigtincan’s best-in-class data privacy and security standards, including certification in SoC2, ISO27001, and CStar. Our products are tested and vetted by leading security tools and penetration testing (more from GRC here, maybe).</p>
<p>To learn more about Bigtincan’s security and data privacy standards, please visit our platform security page and privacy policy at www.bigtincan.com.</p>


<p>Customer data security is a priority. Our AI features always respect established content permissions and access controls. GenieAI only processes content that a user has permission to access, ensuring that responses are based solely on content available to the user.</p>


Artificial Intelligence Governance Charter

FAQs on Bigtincan’s use of AI

Were the tools we use for our products tested for vulnerabilities?

Were the tools tested for biases?

Describe the approach to test the data for bias which may be the result of different data limitations such as missing data, observational intensity, logistical limits, or low participation from different demographics.

Define your bias/fairness testing methodology.

Bias detection and fairness metrics should be pre-identified during the AI/ML Solution design phase to avoid the potential risk of erroneous or unguided measurement.

The pre-identified metrics should be used to repeatedly test and measure the AI/ML Solution performance across different subgroups applicable to the use case.

How was the output evaluated for fairness, including unfair or harmful stereotypes? What were the results of the assessment?

TLS scan results

Vulnerability scan results

Process for confirming accuracy of output

Continuous Monitoring and Improvement: What processes do you have for continuously monitoring and improving the accuracy of your AI services?

Has the AI feature or LLM undergone independent statistical validation for accuracy of its outputs?

Does Bigtincan send my data to 3rd parties for model training?

Do I have a choice in what LLMs I can use?

Will my data be used to create, develop, or improve Al features in the product?

Will product users know when or if they are utilizing an Al-based feature?

Does the product provide alternatives or opt-in/out functionality for Al- enabled features?

What data privacy standards do Bigtincan generative Al features abide by? How are these standards tested?

How do Bigtincan’s generative Al features handle content permissions?