A large manufacturer of consumer products invests in new technology to record, store and analyze customer service calls. A key feature of the technology is that it uses machine learning and other forms of artificial intelligence to help the company track features of its communications with its customers and improve the customer service system. After a few years of operation, the company collects tens of thousands of hours of audio data. The company is sued regarding its consumer practices, and the plaintiffs issue document requests seeking all communications between the company and its customers regarding certain products. The general counsel seeks advice on the company’s obligations to produce the stored audio data and how best to review the data to identify responsive communications while limiting the costs.
Preservation and Discovery of Audio Recordings
Companies are increasingly creating and retaining large quantities of audio data for both internal business purposes and regulatory requirements. For instance, the Securities and Exchange Commission (“SEC”), the Commodity Futures Trading Commission (“CFTC”) and certain other regulatory entities require that certain oral communications—e.g., telephone and “squawk box” conversations—be recorded and maintained for certain categories of employees. Once created and maintained, audio recordings may be discoverable in civil litigation, criminal investigations or other enforcement proceedings.
In the civil context, electronically stored information (“ESI”), including “sound recordings,” is discoverable under Rule 34 of the Federal Rules of Civil Procedure if it is “stored in any medium” from which it can be obtained “either directly or, if necessary, after translation … into a reasonably usable form.” Audio recordings are, thus, treated under the Rules like any other medium, such as email, and may be considered discoverable ESI. Further, requests for audio data in civil discovery are likely, at least initially, to be coextensive with document requests and are, therefore, likely to cover many custodians over a long period of time. In such cases, compliance may be unduly burdensome.
Rule 26 protects a party from being forced to produce ESI from sources that are “not reasonably accessible because of undue burden or cost.” However, there has been little explicit guidance from the courts on the degree to which this rule may apply to requests for large amounts of audio data.
Where a company has invested significant resources to purchase and implement a system to record, store and analyze audio data, a court could (perhaps wrongly) presume that the data is reasonably accessible and that the company can use the analytic power embedded in the company’s own technology to identify responsive documents. While this technology is, in many ways, still in its infancy, it may be difficult to convince a court of the technology’s limitations.
Strategies and Best Practices
As a threshold matter, it is important to know what audio data is being recorded and how long it is being maintained. Typical examples of sources of audio data include:
- Recorded telephone lines: Telephone lines are most often recorded because it is either desired for business purposes or required for regulatory purposes.
- Recorded “squawk boxes”: Squawk boxes are intercom systems used by financial professionals to broadcast offers, bids and other market information to traders and other market participants.
- Voicemail: Voicemail from work or personal cell phones or landlines is a common source of audio data.
The collection, processing, review and production of such sources of audio files have the potential to be costly and time consuming. However, there may be ways to narrow broad requests—through negotiating with counterparties or by conducting preliminary discovery—that are likely to both satisfy the regulator or the court and help reduce the costs and burdens involved. These include:
- Using technology that produces searchable data. Increasingly, e-discovery vendors have the capability to process audio data in ways that render that data searchable (either through speech-to-text or phonetic search methods). Audio data that is searchable can drastically reduce the amount of data that needs to be reviewed in order to fully respond to a request. Counsel should consider identifying vendors with the requisite capabilities if audio data needs to be reviewed and produced.
- Leveraging advanced analytic technology. If the company uses advanced technology to record analyze, and store its audio data, consider whether this technology can be leveraged to limit the volume of data that needs to be processed and reviewed, particularly if the technology has embedded artificial intelligence capabilities. When discussing the use of this technology with the court or opposing counsel, however, it is important to clearly communicate the limitations of the technology early to manage expectations about what the technology can and cannot do in the discovery context.
- Narrowing the scope. Conducting witness interviews and at least a limited review of written communications prior to processing audio data may help narrow the list of individuals at issue and the date ranges and sources of audio that need to be searched. This is important because, even if search terms and analytic tools can help reduce the volume of audio data that requires manual review, processing audio data can be very expensive.
- Reviewing and tailoring retention policies. Where audio is recorded and maintained pursuant to regulatory requirements, there is seldom any reason to maintain the data beyond the regulatory requirements. Where audio is recorded for business purposes, it is likely to become stale after a certain period of time. The company’s retention policies should be reviewed and tailored to meet applicable regulatory requirements and business needs, and careful thought should be given to whether any audio data should be maintained beyond those requirements, with appropriate consideration given to the existence of any litigation holds that might require further retention.
- Instituting a collection protocol. Counsel should also consider instituting a protocol for collecting audio data. Consideration can also be given to whether existing and future litigation hold letters should specifically identify audio data sources on company-issued and personal devices as among the categories of documents and data to be retained, both to ensure that clear instruction is given and to document compliance with preservation obligations.
In conclusion, companies that routinely record and store audio data for business or regulatory purposes may face the prospect of reviewing and producing large quantities of audio data in the regulatory, criminal and civil litigation contexts. The complications and expense associated with these productions demand that companies ensure that they have appropriate systems in place to help effectively and efficiently comply with requests for audio data.