Comment: Video killed the radio star!
This is what The Buggles told us in their number one hit back in 1979. Until then of course, audio had reigned supreme in terms of music on our radios.
However, in terms of sources of evidence and non-written media, in commercial cases at least, video has generally remained on the back-burner, while audio files have featured – be they recordings of trader negotiations, interviews, general phone calls, voicemails, call centres or even recordings of court proceedings.
What next for video?
Things may be about to change. The enforced and swift move from gatherings around meeting room tables to a variety of video-conferencing platforms (notably, but not exclusively, Teams and Zoom) brought about by the global Covid-19 pandemic, means that video has featured prominently and could well become one of the more important medias in terms of evidence in the months that lie ahead.
Whether or not the work from home culture will continue as the world’s lockdown restrictions relax remains to be seen. Perhaps it will become less prevalent than it is currently, but it is highly likely that we will not fully revert to office and in-person based interactions, so the corpus of potential video evidence will continue to grow. Even if it does not, we have plenty of video created over the last few weeks to contend with.
Will video-conferencing recordings be relevant?
DSARs – the fear of an onslaught of Data Subject Access Requests following the introduction of GDPR did not materialise for the most part. However, if we assume that one of the main proponents of DSAR applications is the disgruntled ex-employee, we would do well to acknowledge that since May 2018 when GDPR was introduced, generally, employment markets had been stable and healthy. Now of course, things will change dramatically. We are already seeing redundancies on frighteningly large scales and as furlough schemes come to a conclusion, sadly we might expect many more. Thus, the potential for an increase in disgruntled ex-employees wanting to understand more of how their former employers dealt with redundancy selection processes, consultations and any perceived performance issues may well lead to an increase in DSARs. Much of what ordinarily would have been dealt with in face to face meetings, internal department, board or committee meetings will instead have been dealt with on video conferences.
Litigation/Arbitration and Investigations
Video has of course always been an important source of evidence, but typically there was a narrow use case, and it tended to be more central to the criminal end of the law than the civil.
The rapid shift to video collaboration platforms means that the format is now ubiquitous, and therefore its provenance to dispute is more likely.
Further, the rapid shift in behaviour occasioned by Covid 19 brings obvious risks. Take the advice from the NHS website:
The problem is when short-term becomes long-term, and when specialists in one discipline (say medicine) are suddenly asked to become IT gurus. Digging further into the NHS advice for instance, there is advice that MS Teams has been assured as a secure video conferencing tool, but that Zoom, and WebEx have not. “This does not mean they should not be used, but it is important to note that it is an organisation’s own responsibility to perform risk assessments on any products that are used.” In practice, within the NHS there is a danger that this risk assessment simply becomes a recommendation to look at a link to a government website, so responsibility falls back on the non-IT professionals.
This is in an area where people are acutely alert to the sensitivity of data. In other industries, there is a real danger of user unfamiliarity opening up individuals and companies to all sorts of risk.
For instance, in certain states of the US it is criminal wiretapping to record a telephone conversation without consent.
The problem lies in the fact that it is now much easier to record video, and in some instances it may happen automatically without participants realising. Most platforms do have the necessary security settings, but users may not have had the training to apply them properly. In parallel the platforms are evolving quickly so there is a need for ongoing training and making sure that people upgrade to the latest versions.
In the US there is already platform related litigation, for instance the Zoombombing case where “Zoom’s purported security flaws have allowed “uninvited men” to crash a Texas dance instructor’s virtual burlesque and pole-dancing classes, according to a proposed class action in California federal court, adding to the long list of recent privacy lawsuits aimed at the videoconferencing platform.” Law 360
You could do worse than pay attention to the current advice from the NHS website:
· ensure that you update the software frequently for any tool you use;
· ensure meetings are password protected – otherwise uninvited attendees may be able to join or disrupt your meeting;
· hold people who have joined the conference in a waiting area until you have verified their identity;
· be aware of phishing risks with links/attachments in video chat.
What challenges are peculiar to video?
A significant challenge of video (in eDiscovery terms) is its unfamiliarity. We are used to thinking about email, where it is stored and how it is collected. Above all, we are used to using text for search and analysis. When something is written down, yes it can be ambiguous – but in general there is a precision and concreteness to words. Martin says: “having sat on a jury and watched the same video from three different angles it is astonishing how hard it was to be sure of what had happened. And these were good recordings.” Much video evidence is of poor quality (CC TV cameras) or showing only a partial view. When it comes to video conferences – there may be stray children wandering by, or internet related glitches (but only for some). It is entirely possible, for instance, that a crucial point may not have been heard by a participant in a video call even though they were (from a metadata perspective) clearly attendees and present.
Then there is the challenge of storage. Videos recorded on MS Teams, for instance, are saved to an entirely separate application – Stream. Beyond a certain point, there is an additional charge to store data. Many organisations are clearly going to err on the side of parsimony – resulting in the evidence not being retained.
Conversely, it is increasingly easy to end up with data stored in unexpected places. MS Notes allows you to embed video within it. As for the proliferation of random videos on WhatsApp – well, our advice is don’t let your children take over your phone (though perhaps needs must in lock-down). But more seriously, there is a habit of informal sharing of data, and video can move from platform to platform.
At the more extreme end there is deep fakery. Quite recently this was an interesting outlier. If one were to google the term don’t be surprised if your first hit is: “8 best deepfake apps and websites, you can try for fun.” This is going to get harder.
Then there are governance and ethical questions. Video data is almost by definition personal data, and facial recognition is a key element of any analytics in video. Recently IBM announced “IBM no longer offers general purpose IBM facial recognition or analysis software” because of concerns over privacy and human rights. eDiscovery is not “general purpose” but it does raise a question regarding the basis of such processing.
The biggest challenge however is the familiar one of cost. You can look at a document and immediately understand whether it is likely to be of interest. The visual scan is instant. With video you either have to watch a film in its entirety, or you need to make the decision to invest in the right technology for analysis. Given that this is likely to be expensive, on what basis do you make the decision?
For the moment, it may be possible to avoid engaging with the challenge on the basis of proportionality. But for how long? Should the video record be the only record of a decision or important interaction, then clearly the proportionality argument falls away.
Recording of video conferences
Many video conferences and meetings will have been recorded and all participants will have been made aware of the fact at the outset of the call. We might expect therefore that such recordings are stored and managed in accordance with a company’s routine document management policies. However, what is the position with meetings that were not intentionally recorded or announced as being recorded? What is their legal status? It is likely that a log of the meeting will exist somewhere, but what metadata will the log contain? Who was on the call? Who attended with video – and therefore had access to displayed content – and who joined by audio only? In multi-attendance calls, not all video is displayed on the screen at all times. If a participant is silent, it may be that others on the call do not see that person’s image on screen. However, in terms of sentiment analysis (for example), their appearance and mannerisms might be of relevance. What aspects of the full call are actually recorded, stored and available for analysis? Is the meeting recorded automatically? If not, can the recording be retrieved at all – either readily or by forensic means?
Another consideration is the extent to which video conferences and meetings have been properly and fully minuted. Do video meeting attendees take the same comprehensive notes that they may have taken had the meeting been in person? Even if they did, were those notes properly retained and recorded? If not, the video recording itself will take on far more importance as the record of the meeting.
If video conferencing is to remain a preferred means of communication and media as the world returns to its new normal, companies would do well to make a policy decision that video calls are recorded if it is likely that matters under discussion are likely to lead to any dispute or to form an official record.
Ease of preservation/retrieval
It has already been mentioned that there may be a cost issue here. Video files are large so there is an incentive not to keep them. If they are kept – how are they retrieved and how long does the process of retrieval take? Are there contractual terms in place, especially with cloud providers? Experience recording a demo video not only gave specific information that a 45-minute video is a considerable 3GB of data, but also demonstrated a degradation of quality between the local copy and the version in Stream. If the quality of the copy is important, there might be a need to locate the original version, but this will not have the related chat and associated communication. To take an example from the days of hard copy, this is the equivalent of having a document with marginalia, and one without.
There may well be a need for a two-phase process – analysing email and chat and then searching metadata to isolate the important video content.
A historic issue encountered from the world of audio (trader workstations) is the fact that recording machines are not necessarily set-up in synch, or with the correct time stamps. Engineers are often concerned with facilitating day to day operations, rather than piecing together events from the past. Collection processes may not be standardised – and IT staff given the job of pulling data off systems may do so in their own idiosyncratic ways. It is not unknown for the task to be shared out among a number of staff, who will approach things in slightly different ways, leading to inconsistent data formats. This can be confusing when it comes to eDiscovery, take time to rectify, and therefore add cost. As a rule therefore (as with any non-standard data type) it is important to understand and document the process before you start.
Then there is formatting of the video for ease of use within review platforms. For instance, Nuix Discover works best with MP4. There may be different output options for video. For older video formats there may well be a need to convert.
Audio and video files have their own peculiarities when it comes to the need to redact. Whereas text files lend themselves to relatively easily applied masking or obliterating (either temporary or permanent), audio or video files need to be edited and/or spliced to achieve the same result. Added complications when removing sections of recorded content (as opposed to masking it) are that the time stamps will be out of synch with the master (unedited) version.
The uptake of video-conferencing to allow business to continue during lockdown has led to a new phenomenon in terms of redaction, GDPR etc. What does the holder of the recorded file need to do to redact the unscheduled appearance of an attendee’s partner or child during a meeting – either in camera shot or audible in the background? What of an attendee’s location view? What personal information might that portray?
The good thing is that there are tools that facilitate audio/video redaction, but knowing exactly where in the recording they need to be applied is time consuming and expensive. Also, it does not follow that an entire section of the file can be redacted. For example, if an attendee’s child comes into view of the camera at the very point a highly critical discussion is taking place, the audio segment must be retained and only the video element should be redacted.
There are a number of processing and analytics tools out there specifically aimed at the eDiscovery sector, notably Veritone and Intelligent Voice, which turn speech to text, synch with the audio and automate speaker separation. This can be an effective way to isolate out key custodians, and locate to whom they are speaking.
For production purposes, the tools allow for the simultaneous redaction of voice and text. In the criminal context for example, it is common to go further – for instance identifying and isolating heads (which can be blurred out) number plates, brands of watches – all sorts of personal identifiers. But powerful though these are, none of them are perfect.
There is always (as with conventional documents) going to be a step where a human eye needs to review, check and manually edit.
Companies with large amounts of video data may well have tools of their own. As always, our advice is to explore whether they can use what they already have.
Analysis and review methodology
As ever, there are many ways to analyse audio and video data and there is no right or wrong way (well there are probably many wrong ways!). One method that has served well in terms of some audio review is to convert speech to text and then review the text. There are obvious pitfalls and opportunities to miss critical data relying on this method alone, but it may well be a rough and ready way to identify passages of content that might merit closer attention. Also, with accompanying chat data, some have elected to display the content in spreadsheet form, which might work well for the case team, but is not an attractive or intuitive way of putting the evidence to a witness for comment.
If the video is to be reviewed in its native (or similar) format, it is likely that the review team will need specialist equipment or third party plug-ins, keeping hands free to type contemporaneously as appropriate. The problems increase where elements of speech may be inaudible or where participants speak over each other.
Of course, a very powerful means of achieving an informed review of video data, is to synch it with any accompanying chat data (whether from the same programme, or even a different programme or app).
As Panoram founder and Director Greg Wildisen rightly points out, it is useful to draw a distinction between “speed to review” and “speed of review”. The latter can be particularly expensive. Due to the nature of the media, it is not unreasonable to expect to apply a ratio of 4:1 in terms of review time. That is four hours review time per one hour of recording. The former relates to what processes are necessary from the moment the recording is identified and extracted to the point at which it is rendered available for review in a suitable format. This might include marrying it with any relevant chat data, as suggested above.
Whilst always considering how long the review team will need to complete the task, it is good practice not to cut corners when considering the speed to review. This is the opportunity to consider pre-review analysis and to apply thought and tools to hone in on the segments of recording that merit full-on review and perhaps apply a staged approach to identify which areas require only a “gist” review. Some sections might be excluded from review altogether
eDiscovery has always been largely about bringing structure to unstructured data and finding a thorough, defensible, and proportionate way to get to and produce the important evidence. Video data is no different, it is just that it has rather suddenly gone from being at the fringe of things, to being potentially central.
This article raises a lot of questions, for which the answers may not as yet be perfect and the solutions by necessity somewhat improvised. As virtual working becomes routine – the approach to video eDiscovery will no doubt become more standardised and more automated. In the meantime though, a pragmatic and non-fearful approach makes sense. Good data governance and security policy is clearly going to make life easier. Understanding how data is stored and documenting / standardising any collection approach is key. Use related data to identify which video data is central. This will give you a proportional approach. Should you need clever analytics and AI – then the tools are available.
Watch this space!
Vince Neicho: Vince joined Integreon in June 2017 as an expert legal solutions consultant, with a focus on law firms and corporate legal departments engaging in e-disclosure, e-discovery and document review. Prior to joining Integreon, he was with Magic Circle law firm Allen & Overy for 42 years, where he set up and led their global litigation support / eDiscovery teams. In this role, he assisted key clients from many sectors with the design of innovative, defensible systems, processes and solutions to meet their discovery and document management obligations.
Martin Bonney: Martin is head of managed services/eDiscovery at Panoram Digital. He has worked in the eDiscovery field for over 25 years – first for Linklaters and then he founded the London office of nMatrix, which became Epiq where for many years he headed up the Consulting Team. eDiscovery is by its nature all about bridging the worlds of law and digital – with a need to deal practically with the past but plan effectively for the future. Martin believes that to make the most of our growing digital capabilities – we need to maximize the potential of existing platforms (notably Microsoft O365) giving lawyers the tools to boost their own efficiency, control costs and deliver the best results for their clients.