Using AI notetakers and scribes in your online meetings - issues and guardrails

The use of AI notetakers and AI scribes in the context of community, human, and health services has come up as an area of excitement and concern. Many people are starting to use them and there are concerns about privacy, control over meeting notes and artefacts. Many have been using them without having any policies in place. There is a desire to have some useful and practical guidelines and policies in place.

In this article I'm providing some useful articles and sources that I have found to provide an overview of the issues that service providers should be aware of. Below the overview of issues to be aware of I've provided some high level policy considerations generated by AI itself.

And, no, this is clearly not legal advice. There's some of that in the article. AI notetakers are easy to set up and bring or send to a meeting. But there are many things to consider below the surface before you used these tools in your work. This article is simply providing links to authoritative sources I've found (there are undoubtedly many, many others) and useful tips to help you start the conversation in your organization and to help guide your responsible use of these useful tools.

First, some useful sources I've come across:

Artificial Intelligence - Protecting Privilege with Artificial Intelligence in the (Virtual) Room
"Recent developments in Artificial Intelligence (“AI”) have brought the mainstream use of AI into our daily lives and daily work lives. While AI has been used by litigators for over a decade in e-discovery, new applications and tools that utilize AI are prompting new questions. One of these questions is: whether and how organizations can use AI without waiving privilege. Many meetings, including with legal counsel, are now conducted using videoconferencing platforms such as Microsoft Teams, Google Meet or Zoom. AI note-taking applications often ‘attend’ these meetings to transcribe meeting notes and generate meeting summaries. Does the presence of such an AI notetaker during a meeting result in waiving solicitor-client privilege? To answer this, we need to consider both privilege and how AI works."

AI Notetakers – Evaluating the Risks Along with the Benefits
"The explosion of generative AI has spawned a wide range of personal and professional tools and applications. One noteworthy (no pun intended) example of those tools and applications is notetakers that can capture, transcribe, and organize the content discussed at meetings (virtual or otherwise), enabling participants to more meaningfully participate in the meeting/discussion. They can even enable an individual to not be present at the meeting at all and not miss out! Of course, like any new AI or other technology, it is important to consider the risks along with the benefits."

AI Meeting Notetakers: Navigating Privacy and Security
"In an era where artificial intelligence (AI) seamlessly integrates into our daily workflows, AI notetakers in online meetings have become a new frontier for innovation. As an (MSP) specializing in Cybersecurity for regulated industries, we've observed an ever-increasing uptick in deploying these tools during professional interactions. While these AI notetakers promise efficiency and improved productivity, they also bring forth significant concerns regarding privacy and cybersecurity."

AI Summarization: The Easiest and Fastest Win for Gen AI
"Since the advent of CliffsNotes, we’ve recognized the value of summarizing documents. AI summarization takes this concept to new heights, using sophisticated algorithms to distill large volumes of text into concise, easily digestible summaries. It replaces lengthy reports and manuals with bite-sized summaries that highlight key points. Need to quickly understand a research paper? Summarization delivers the essence in no time. That’s the power of AI!"

And a useful report looking at how "Artificial intelligence (AI) scribes and robotic process automation (RPA) have emerged as potential solutions to address these challenges, and this project aimed to evaluate their impact on primary care and potential benefits for the Ontario healthcare system."

And now, an AI-generated overview of the key issues and a high-level policy regarding the use of AI notetakers and AI scribes in community, human, and health services (this link takes you to the Perplexity.ai page where you can see my prompt as well as the sources it used for the information below, which I have lightly edited and also added some points from the article AI Notetakers – Evaluating the Risks Along with the Benefits, added in quotes):

Overview of Key Issues

Privacy Concerns

One of the primary issues with using AI notetakers and scribes is the potential impact on privacy. Some key privacy considerations include:

• Ensuring AI tools comply with relevant privacy laws and regulations like PHIPA, PIPEDA and any relevant privacy laws in your province and jurisdiction.
• Obtaining informed consent from patients/clients before using AI scribes, including disclosing if data will be used to improve AI models.
• Verifying AI vendors follow proper security standards and access controls to protect sensitive data. Do you know how your data is being used, accessed, or shared by the AI vendor you're using? You should.
• Conducting privacy impact assessments before implementing new AI tools.
• Are you sharing a link to the AI notes with all participants? How will you ensure they will not be shared more widely if that is a priority?
• How are you controlling or moderating the use of AI notetakers being sent by meeting participants in your meeting? What framework do you have in place to allow or deny these external notetakers being present and recording your meeting?

Data Security

In addition to privacy, the security of data captured and processed by AI is a major concern. Organizations should:

• Ensure AI vendors use encryption, secure cloud storage, and other safeguards to protect data.
• Limit access to AI-generated notes and transcripts to authorized personnel only.
• Have policies for securely storing or destroying recordings used to generate notes, including permanent deletion from cloud-based AI vendors.
• "Is your data used to train the notetaking tool? Some notetaking tools will use the transcriptions generated by customers to help improve the accuracy of the product. Of course, the organization using the tool will need to consider the confidentiality, privacy, and security of the information it permits its notetaking vendor to acquire for this purpose, and whether this practice raises regulatory or contractual issues. The tool might provide an opt out from this use and the organization will want to make sure to train employees to opt out, as needed."
• "What kind of confidential and personal information do you anticipate will be captured by the tool? As with many AI applications, it is critical to understand the use cases that you anticipate being served by the technology. The use cases can be wide-ranging and will be shaped by, among other things, the type of business and activities engaged in, which departments/employees in the organization are using the tool, and other factors. For example, in a law firm environment, using a notetaker likely will raise attorney-client privilege issues. In a healthcare environment, it is likely that a notetaker could capture protected health information (PHI) of patients. However, if a health system’s marketing department is using a notetaker, capturing PHI might be less likely, but still possible. So, when thinking about how your organization will use a notetaker, it is important to consider not only your organization’s regulatory environment, but also who in the organization will be permitted to use the technology and for what purpose(s), what representations have been made about disclosures of confidential and personal information, etc."
• If used for official meeting or client records, how will notes from the AI notetaker be summarized or entered in official minutes or client tracking systems, and will full transcripts be needed to be securely stored for any length of time?

Accuracy and Bias

The accuracy of AI-generated notes is another potential issue, as AI may misinterpret nuanced or complex information. There are also risks of AI perpetuating biases present in training data. To mitigate this:

• Providers must carefully review AI notes to verify accuracy before saving to official records. AI notetakers are merely tools. Humans still need to be involved to finalize and sign off on the accuracy of notes.
• AI tools should be evaluated for bias and fairness, particularly for underserved populations. That's clearly a consideration in the Newcomer-serving sector. And when using AI notetakers with any external clients, everything related to data security and privacy concerns outlined above and in other resources is essential to keep in mind. You should not be sharing client names, personal information, etc., with unvetted AI vendors.

Informed Consent

Obtaining proper informed consent is critical before using AI notetakers or scribes. This includes:

• Notifying all meeting participants that AI tools will be used and obtaining their consent
• Explaining how recordings and data will be used, stored, and potentially de-identified
• Respecting participant choices if they do not wish to be recorded

Impacts on Participation

The use of AI notetakers may make some meeting participants less comfortable speaking up or sharing sensitive information. Organizations should be mindful of this impact.

High-Level Guidlines and Protocols for AI Notetakers and Scribes

From the AI Notetakers – Evaluating the Risks Along with the Benefits article, important questions you need to ask, followed by Perplexity.ai' suggestions for policy considerations.

"Do we need a policy? New technologies like generative AI and their various iterations often raise many questions concerning use in organizations. Indeed, many organizations have adopted policies to guide employees when using another popular application of generative AI technology – ChatGPT and similar tools. Policies can be helpful to establish guiding principles and requirements for employees, such as:

• which notetaker(s) have been vetted by the organization and are approved for use in the course of employment,
• which employees are permitted to use the notetaker and for what purposes,
• guidelines for providing notice, consent, etc.,
• what safeguards should be followed for securing transcriptions with confidential and personal information,
• guidelines for limiting access to transcriptions,
• record retention and litigation hold requirements, and
• how to handle meetings intended to be privileged."

Some policy guidelines:

1. Carefully evaluate AI vendors to ensure they meet all relevant privacy, security, and compliance requirements. Involve legal, IT, and other key stakeholders in this review. Implement end-to-end encryption and secure storage for all data handled by AI notetakers. Ensure compliance with relevant data protection regulations (e.g., GDPR, HIPAA). Choose vendors that comply with privacy laws and have a strong track record in data security.
2. Always notify participants if an AI notetaker or scribe will be used in a meeting or patient encounter. Clearly inform participants about how their data will be used and stored. Obtain their explicit informed consent and respect their choice if they decline.
3. Pause AI notetaking or scribing when discussing sensitive, confidential, or privileged information.
4. Providers must review all AI-generated notes and transcripts for accuracy before saving them to official records or acting upon them.
5. Securely store any recordings used to generate notes for the minimum required retention period. Destroy recordings securely once notes are finalized.
6. Restrict access to AI-generated notes and transcripts to authorized personnel and use encryption and other technical safeguards to protect the data.
7. Conduct regular audits and reviews of AI notetaking systems to monitor for accuracy, bias, and any unintended consequences. Conduct regular security audits and vulnerability assessments.
8. Educate staff and participants about the ethical and legal implications of using AI notetakers. Provide training to all personnel on the proper use of AI notetakers and scribes and the organization's policies regarding their use.
9. Have clear processes for participants to request corrections to AI-generated notes or to withdraw their consent for AI notetaking.
10. Stay informed of the evolving legal and ethical landscape around AI use and update policies and practices as needed to remain compliant and manage risk. Adjust policies and practices based on feedback and technological advancements.

Implementation

• Develop detailed procedures for implementing AI notetakers and scribes in line with your policy.
• Establish an internal review committee made up of a diversity of staff and include client representation to oversee the use of AI notetakers and scribes and ensure compliance with this policy.

Review and Revision

• Review your policy annually or as needed to reflect changes in technology, legislation, or best practices.
• Revise your policy as necessary to ensure it remains effective and relevant.

In summary, while AI notetakers and scribes offer benefits in efficiency and accuracy, organizations must proactively address issues of privacy, security, consent, bias and accuracy. Having clear policies that put patient/client rights first is essential for the responsible adoption of these tools in community, human and health services. Ongoing monitoring and refinement will be needed as regulations and best practices in this space continue to evolve.