The messages were set to delete within an hour, and some of those settings were changed after a litigation hold had already landed. That was really the whole idea.

If you practice long enough, you learn that the tools change but human nature doesn’t. Caveguy Ugg didn’t want anyone to know where he buried the mammoth haunch. Today the digging stick is a setting inside Signal, and the dirt is a “disappearing messages” timer. Same instinct, better encryption.

A million years post-Ugg, the same instinct turns up in the Delaware Court of Chancery. It’s a decision worth reading once, and worth repeating to a client twice.

Wait, isn’t the whole point of an encrypted app that the messages are gone?

Gone from the phone, maybe. Not gone from your obligations.

In In re World Wrestling Entertainment, Inc. Merger Litigation (Del. Ch., May 2026), Vice Chancellor Laster addressed exactly this. Senior people ran Signal with auto-delete turned on, after receiving litigation holds, several manually changed individual chats so messages would disappear in a little less than an hour. The court didn’t shrug. The court held that the first hold already reached Signal communications that overlapped with the potential deal. And even apart from the hold, it found that the duty to preserve sale-related evidence had arisen by August 31, 2022. That was months before the later sale-process hold. Sophisticated parties in that situation should reasonably have anticipated litigation.

The line worth taping to your monitor: litigation holds are not self-executing. Sending the memo, and even making sure they got it, is not the same as preserving the data. Somebody has to identify the relevant apps and devices, verify the settings, and actually disable the auto-deleted function.

So, what did it cost them?

The court didn’t reach for the biggest hammer. Rather than make a default judgment or give a blanket “assume the worst” instruction to a jury, it did something more surgical. It was arguably more dangerous for the sanctioned side: it presumed certain specific facts to be true and then shifted the burden, requiring the defense to overcome those presumptions by clear and convincing evidence rather than the usual preponderance. A bit of a switcheroo on the usual burden of proof ground floor.

Read that again if you litigate. The evidentiary floor moved. That’s not a slap on the wrist.

Now, I’m a forensic examiner, not a lawyer, and the doctrine here (when the duty attaches, what sanction fits) is your department, and it varies by jurisdiction. What I can tell you is the part that lives on my side of the table: how this actually plays out in the data.

Can’t your people just recover the deleted Signal messages?

Sometimes, but not always, and not the way Professor Google’s many advisors say. Here’s what’s real.

Disappearing-message apps are built to leave little behind, but “little” is not “nothing.” Copies and traces have a way of surviving in places people forget. A forensic acquisition or Signal backup made while the data still existed. The conversation on another participant’s or linked device where some data has survived. A screenshot somebody took. Notification artifacts preserved elsewhere. A message quoted in an email. Encryption is a strong lock on the message itself. It does nothing about the copies that already walked out the door and went for a jog.

But if the content is truly gone, the forensics may not bring back the words. This matters much for setting expectations with a client. What may be able to show is the activity on the device or app artifacts: that an app was installed, that auto-delete was switched on, that a timer was shortened at a telling moment. Note that we’re talking about what the device did, not who was holding it. Putting a specific person at the keyboard is a separate and much harder question. But in a spoliation fight, showing that a message was set to vanish, and when the setting changed, can matter as much as the words that vanished.

So, what to do?

If you represent the party with the duty, the practical steps are unglamorous and they work.

Identify the apps early. Ask, in plain language, what everyone messages on — not just email and texts. Signal, WhatsApp, Snapchat, Teams, the group chat nobody mentions.

Turn off disappearing settings the moment litigation is reasonably foreseeable, and confirm that preservation is actually occurring rather than trusting that the memo did the job.

Preserve any available backups, forensic images, and secondary devices before data rolls off or is overwritten. But don’t assume an ordinary phone backup contains Signal history—or that a Signal backup captured short-lived disappearing messages. It’s a good idea to collect that before today’s backup writes over yesterday’s.

Document your preservation, not just your hold. When someone later asks what you did, “we sent a notice” is a weaker answer than “we sent a notice and confirmed the settings on every device by this date.”

None of this requires a particular level of paranoia. It requires remembering that a message built to vanish can still leave a record of having been built to vanish.

Ever had a “disappearing” conversation turn out not to be the dead end everyone assumed? I’d like to hear how it surfaced.

Steve Burgess is a digital forensics expert witness with more than 40 years of experience and over 20,000 devices and digital media examined. He is the principal of Burgess Forensics, founded in 1985.

Burgess Forensics: (866) 345-3345 | steve@burgessforensics.com

Don’t miss a single issue of our informative newsletter … Subscribe now!

The post When “Auto-Delete” Becomes “Adverse Inference” appeared first on Burgess Forensics.

“It’s encrypted, so I guess we’re just out of luck.” I hear some version of that from attorneys and other clients more often than you’d think, usually said with a kind of resigned finality, as though the phone in evidence had sealed itself inside a block of concrete. Sometimes it’s true (well, not the part about the concrete). More often, it isn’t — and treating encryption as the end of the inquiry leaves evidence on the table that the other side may be perfectly happy to collect.

Here’s the thing worth understanding: encryption locks the front door. It rarely locks the windows, the copies, and the spare key under the mat.

Modern devices encrypt data at rest. When an iPhone or a current Android phone or a laptop running FileVault or BitLocker is powered off or locked, the data on it is scrambled and, without the key, is effectively unreadable. That part is real, and it’s strong – it’s real strong. But “the data on that specific device, while it is locked” is a much narrower thing than “the information you’re after,” and the gap between those two is where most access actually happens.

The passcode is the whole ballgame — and people are careless with it. Full-disk encryption is only as strong as the credential protecting it. People reuse passwords, write them on sticky notes, store them in a notes app, share them with a spouse or an assistant, or pick something guessable. When access to a device is lawfully available — through consent, a cooperating party, or a court order — the encryption stops being an obstacle the moment the passcode is in hand.

And by the way, the data usually lives in more than one place. This is the point often missed. A message exists on the sender’s phone and the recipient’s. When the message’ sender deleted it from their phone, it doesn’t do anything to the recipient’s copy of the message. Photos sync to iCloud or Google Photos. Documents sync to cloud storage. Phones back up to a computer or to the cloud, sometimes automatically, sometimes in a form far easier to access than the locked handset itself. The encrypted device in the evidence bag may be the hardest copy of the data to reach — and the least necessary, once you map where else that same information lives.

Cloud accounts are their own doorway. When the data has synced to a provider, the relevant credential may be an account password rather than a device passcode, and the legal path may be a subpoena or warrant to the provider rather than an attempt on the hardware. Different lock, different key, often a more productive one.

The law here is genuinely unsettled, and that’s your department, not mine. Whether a person can be compelled to disclose or enter a passcode touches the Fifth Amendment, and courts around the country have landed in different places on it — some applying a “foregone conclusion” rationale, others declining to. Compelled use of a fingerprint or face has its own tangled line of cases. I’m a forensic examiner, not a lawyer, and I won’t pretend the doctrine is settled when it plainly isn’t. But knowing that these avenues exist — and that they’re contested — is the difference between assuming a device is unreachable and asking the right questions about how it might lawfully be reached.

So what should you actually do? Stop treating “it’s encrypted” as the end of the road, as a verdict. Treat it as one locked door in a building with several entrances. Ask where else the data lives — the other party’s device, the cloud, a backup, a synced computer. Preserve all of it early, before someone decides to tidy up. And bring in a forensic examiner before you conclude anything is impossible, because what’s feasible depends heavily on the specific device, the operating system version, and how the data was stored — details that change constantly and that a competent examiner tracks for a living.

Encryption is a strong lock. It is not a force field. The attorneys who understand the difference get to the evidence; the ones who don’t talk themselves out of it.

Have you ever had a case when a locked or encrypted device in your case was or was not actually a dead end? I’d love to hear about it.

Don’t miss a single issue of our informative newsletter … Subscribe now

The post Encrypted Doesn’t Mean Untouchable: What Attorneys Should Know About Device Access appeared first on Burgess Forensics.

You thought you sent a one-page letter. What you actually sent was a one-page letter and a small pile of paperwork the letter filled out about itself when you weren’t looking.

That paperwork is metadata — data about data. And it travels with your files whether or not you invited it along for the trip.

Metadata is the stuff a document quietly jots down while you’re jotting down words. There’s a lot of potential information there: Who created it, when, and on what computer. Who edited it, and when they last saved it. Where a photo was taken, down to the GPS coordinates. What camera, what settings, what software. None of it shows up on the page. All of it comes along, like a stowaway.

Most of the time this is harmless, but occasionally, it’s the whole story.

A photo is a very talkative little file.

Take a picture with your phone and you’ve created a small autobiography. The image, surely, but tucked inside is a section called EXIF data: the make and model of the phone, the date and time down to the second, and, if location services were on, the exact spot on Earth where you stood. Share that photo in its original form and you may be handing over your home address without meaning to.

The good news: most social platforms started stripping this out a few years ago when there was a public hue and cry about it. The bad news: “most” is not “all,” and emailing the original file, or dropping it in a shared folder, sends the whole talkative package along.

Documents keep a diary too

A Word document remembers more than the final draft. Depending on your settings, it can carry the author’s name, the company the software was registered to, how long the file was open, and sometimes, a list of former edits and tracked changes and comments you thought you’d removed. Every “on second thought, delete that paragraph” can live on in the file’s memory.

The Internet is dotted with cautionary tales of press releases and legal filings that were sent out with the edits still readable underneath. The same is true with newsworthy congressional hearings. Most of us are not popular enough to warrant interest from the government. Still, the words on the page said one thing but the metadata said “here’s what we almost admitted.” Oopsie.

So, what to do?

There’s no need to get paranoid about this. Depending on what you’re shopping for, your grocery list is not a national secret. But a few practical habits go a long way:

• Before sending anything sensitive, look under the hood. In Word on Windows, “Inspect Document” should find and remove hidden data, comments, and tracked changes. Do it on the final version, not the draft.
• Turn off location tagging for your camera if you don’t need it — or scrub EXIF data from photos before sharing the originals. By the way, if litigation is foreseen that involves these photos, don’t scrub the EXIF metadata. It will be considered spoliation of data and will go poorly for you if and when it goes to court.
• Remember that “delete” inside a file often just means “hide.” Removing a comment from view is not always the same as removing it from the file.

None of this requires becoming a hermit. It’s the digital equivalent of checking your pockets before you send the coat to the cleaners.

Metadata isn’t sinister. It’s just honest — sometimes more honest than we’d like. The trick is knowing it’s there, so you decide what to share instead of the file deciding for you.

Steve Burgess is a digital forensics expert witness with more than 40 years of experience and over 20,000 devices and digital media examined. He is the principal of Burgess Forensics, founded in 1984.

Don’t miss a single issue of our informative newsletter … Subscribe now!

The post The Metadata You Didn’t Know You Were Sending appeared first on Burgess Forensics.

The format language in a discovery request is easy to skip over. It looks like boilerplate. It reads like boilerplate. It is actually boilerplate. And so it gets waved through: “produce as PDF, that’s fine.” Maybe not so fine when that same attorney later pays me to explain why the file on my screen can’t answer the question the case now turns on.

Here’s the thing the other side already knows: whoever picks the format picks what you get to see. It’s just how the rules work. Under Federal Rule of Civil Procedure 34(b)(2)(E), the party asking for the documents gets to specify the form they arrive in. Don’t specify, and the choice falls to the producing side, with “reasonably usable” form as the only floor — and reasonably usable is a long way from native. Say nothing about format and you’ve handed them the pen.

Native files vs. PDFs sounds like an argument for the IT department, right? It isn’t. It’s one of the quietest, most consequential fights in the whole discovery process, and it’s worth having on purpose.

A PDF is a photograph of a document, not the document

A native file is the thing itself – the spreadsheet with its formulas still doing math, the email with its full routing header, the Word file that still remembers every draft, the photo that quietly wrote down where and when it was taken. I have a paragraph in my engagement letter saying so, albeit a bit more succinctly. A PDF or a TIFF is a picture of that file after someone chose the pose.

What really gets me is when evidence is produced as a printout of a PDF of a scan of a PDF created from the actual original file. Happens all the time. But it’s several steps away from what really happened and far from the metadata that tells the real story.

Such “pictures” drop exactly the parts that tend to win cases. Most of the metadata, describing fields such as who made it, when, on what device, and sometimes where tends to get lost when a PDF is generated from it. This is the difference between “he says he wrote it in March” and knowing, to the minute, that he didn’t.

A spreadsheet may show you a very precise value of 4,203,722.46 in the flattened PDF version but hides the formulae and structure that generated a possibly different number. A document that previously contained all the things people put in a document before they remembered other people would read the tracked changes, comments, hidden rows? Native format keeps them. The flattened copy forgets them, conveniently.

Hash values and container data are integrity signals that may let you and me confirm that the document produced is the same as the original are casualties of a PDF export. It snaps that verification thread and asks you to just take everybody’s word for it instead.

None of it comes back, either. You can’t un-flatten a PDF into the original any more than you can un-fry a flapjack. Once it’s produced that way, the missing data isn’t hiding from you. It’s gone.

I had a case where the “date digitized” EXIF metadata from a series of photos supported one side’s story. However, deeper in the metadata, there was satellite data that cast real doubt on the story being told. Satellites and their atomic clocks don’t lie about the time or the day.

Why the other side is so agreeable about it

Producing in PDF usually isn’t laziness. It’s control wearing the costume of laziness. A flattened production is easier to redact, easier to Bates-stamp, and, in a happy coincidence, easier to sanitize. The timestamp that wrecks the timeline, the author who wasn’t supposed to be within a mile of that memo, the formula that shows how a number really got built: all of it vanishes in the conversion, and it vanishes wearing the respectable coat of “routine document handling” rather than the orange jumpsuit of spoliation.

To be fair, imaged production isn’t always a dodge. Sometimes there are honest reasons for it — privilege review, redacting personal or protected information, or real proportionality under Rule 26(b)(1) when native would cost more than the whole dispute is worth. Those reasons are legitimate. The trick is to make the other side say so out loud, in the protocol, so that “we imaged it” has to be justified rather than simply assumed.

I should be clear that “save as PDF” is also a genuinely lousy way to strip metadata even when someone’s trying to be honest. It leaves things behind and takes things it shouldn’t. But that’s a different article. For today: the party that controls the format controls the evidence, and they know it even if your side doesn’t.

The fight is won in the ESI protocol, not in a motion six months later.

This is the part I most want attorneys to hear. By the time you’re standing in front of a judge complaining about a bad production, you’re asking to redo work the other side has every incentive to slow-walk into the next fiscal year — and you’re explaining to your client why the schedule, and the bill, went sideways. Nobody enjoys that conversation. Least of all the person who once said “PDF is fine.”

So specify the format on the front end, in writing:

• Ask for native formats, with metadata, by default for anything data-rich — spreadsheets, databases, structured exports — with load files that actually carry the field data. Boring to negotiate. Priceless to have.
• Name the metadata fields you want. Custodian, author, created and modified dates, an MD5 or SHA-1 hash, and parent/child relationships so attachments stay tied to their emails. “With metadata” is an invitation for the other side to read the phrase as generously as their conscience allows.
• Reserve the right to request native format for anything produced as an image, and say so up front — so when you invoke it, it’s a term of the deal and not an ambush they get to act wounded about.
• Bring your examiner in before the language is set, not after the production disappoints. Format specs written without technical input have a real talent for asking, very precisely, for the wrong thing.

The bottom line

This is not enhance-the-reflection-in-the-sunglasses forensics. It’s plumbing. It’s unglamorous, it’s easy to skip, and it is very often the only thing standing between evidence you can build a case on and a tidy picture of evidence you have to take on faith.

Courts increasingly expect native production where format carries meaning, and (hopefully) increasingly treat “well, we already gave you a PDF” as the weak answer it is. But you don’t drift there by luck. You get there by treating format as a substantive term of the case — argued with the same seriousness as scope and custodians, and about a thousand times more attention than it usually gets.

Have the fight early. It’s a great deal cheaper than the one you’ll have later, in front of a judge, with your client and their checkbook watching.

What’s the worst production-format surprise you’ve run into — native you wish you’d demanded, a PDF that turned out to be hiding the whole case, or something else entirely?

Steve Burgess is a digital forensics expert witness with more than 40 years of experience and over 20,000 devices and digital media examined. He is the principal of Burgess Forensics, founded in 1984.

Don’t miss a single issue of our informative newsletter … Subscribe now!

The post Native Files vs. PDFs: Why Discovery Format Fights Are Worth Having appeared first on Burgess Forensics.

Twenty years ago, a forensic report about a hard drive was mostly an argument about whether a file existed and when it was last touched. And of course, trying to recover deleted material. Today the same report might need to address whether a video is real, whether a document was generated by a language model, whether a “deleted” text message ever really existed on the device in the first place, and especially, whether the tool used to answer any of those questions is itself reliable enough to stand behind in front of a jury.

It doesn’t come up in every case, but when it does, we’d better be ready to answer. Digital forensics has always rested on the idea that a method can be explained, tested, and challenged — that’s pretty much the whole premise behind Daubert and Frye. AI-assisted analysis complicates that premise, because a growing number of detection and authentication tools now involve models whose internal reasoning isn’t fully visible even to the people who built them. Courts are already grappling with what that means for the right to confront the basis of evidence against you, and there isn’t yet a settled answer For instance, I’ve seen questions about whether Cellebrite might mis‑label a recovered file as‘deleted’ or treat an active file as if ithad been deleted, even though the platform is historically solid and time‑tested in most other respects. Still, that’s not a hypothetical academic question; it’s a live one working its way through appellate opinions right now and it’s going to shape how expert reports get written well into the future.

What’s changing on the ground, in the meantime, is the volume and variety of source data an expert has to account for. A phone used to mean call logs and texts. Now it means cloud-synced photo libraries, health data, smart-home integration logs, location history from a dozen apps that nobody remembers granting permission to (or maybe that an update changed the permissions you did or did not grant, and metadata trails that span devices the owner may not even still possess. The expert’s job isn’t just extraction anymore — it’s building a coherent, defensible narrative out of data that lives in more places than any one device.

The next several years will bring a few concrete shifts. Although there will certainly be changing laws for changing environments, standards bodies and courts will move, slowly and , toward requiring more explicit validation testimony for AI-assisted tools. Not just “the software said so,” but documented error rates, testing methodology, and version-specific behavior, the same rigor that’s long been expected of DNA analysis and toxicology. Authentication of video and audio is going to become its own specialized sub-field, distinct from general digital forensics, the

way arson investigation split off from general fire science. And attorneys are going to need to get comfortable asking experts pointed questions about tool provenance — not because they distrust the expert, but because opposing counsel increasingly will.

None of this replaces the fundamentals. A well-documented chain of custody still matters. So does a methodology that can be explained in plain English to twelve people, a judge, and possibly an attorney who’ve never heard of a hash value, and an expert willing to say plainly what the evidence does and doesn’t show — that hasn’t changed and I don’t expect it to. What’s changing is the amount of homework required to get there, and how much of that homework now involves tools that didn’t exist five years ago.I’d like to makethis the first in a short series looking at where this field is actually headed, drawing on conversations with people building the tools and writing the standards rather than just my own two cents.

If there’s a specific angle—AI detection reliability, Confrontation Clause questions, how courts are handling deepfake authentication—you’d like covered first, I’m glad to hear it. Trial lawyers and judges who live with these issues every day are exactly who I’m hoping to hear from, so please let me know.

Steve Burgess is a digital forensics expert witness with more than 40 years of experience and over 20,000 devices and digital media examined. He is the principal of Burgess Forensics, founded in 1984.

Don’t miss a single issue of our informative newsletter … Subscribe!

The post The Future of Expert Testimony in a Digital World appeared first on Burgess Forensics.

AI and Elder Abuse, copyright 20025, Steve Burgess

The news is full of AI (Artificial Intelligence) stories. How will it empower us in our jobs? Whose job will it take next? Is it creating actual fake news? 

While there’s a lot of “we’ll see” in the answers to these questions, one immediate and quite pressing issue is how AI is contributing to Elder Abuse. Elder abuse is an action(s) in any relationship where there is an expectation of trust that causes harm or distress to an older person.

American elders were conned out of more than three $3 billion dollars in 2023 in a range of financial crimes, according to the FBI data. AI can make these scammers seem more believable partially because these programs help with content creation and cleaning up human error that might otherwise warn us we’re being served up something fraudulent.

Already, there are so many scams via email, websites, phone calls, advertisements, and offers of romance that there’s plenty to worry about. But does the emergence of AI in everyday life and writing supercharge these efforts to relieve the older amongst us of their carefully husbanded and/or limited resources?

Already, individuals are reaching through our computers and phones to lighten our wallets and bank accounts. There are websites that look like our bank sites, phishing emails that look like they’re from someone or from an institution that we know and trust, ads galore that tempt the reader with promised services, goods, or fortunes.

Already, we give away so much of our privacy in return for the freedom to browse the web and to use other online services. We also regularly share our private information, favorite colors, cars, dates and faces on publicly available social media. But now, there are tools in the guise of fun apps that offer to clean up our photos, or ones to let us have our own voices sing our favorite songs perfectly, our own tin ears notwithstanding.

I see ads every day inviting me to seamlessly insert my voice instead of the original voice – say Sting’s or Sinatra’s, into a popular song. Like karaoke on steroids that we can save and play for or send to others. It’s likely to be fun, but then your voice is being stored in records you have no control over. Thousands of people’s voices are thereby being harvested and stored by some company you don’t know, that has a privacy policy that most of us never read through, and which can be used by these unknown persons in any which way. 

Even if the company offering these services is completely on the up and up, they are a prime target to be hacked by those who would love to use these stored audio clips to clone your voice into realistic but invented audio sent to you via email, or even a phone call.

So what happens when you hear from a beloved grandchild – in their own voices – that they are suffering, or need help out of a jam? And what if you, as a parent, grandparent, or even just a friend, hears from another trusted friend or loved one that they have “such a deal?”

Why, for the cost of nothing at all, you can use chat GPT right now to create a picture of you standing next to the President or your favorite rock star. Some generative AI tools can even create  a video of the two of you singing 70’s rock anthems together.

But nearly as easily, a bad actor could create a picture or audio of your child or friend, needing bail, stuck without airfare far away, injured, in a foreign prison cell, or in the hands of dangerous thugs, begging for your help. 

As technology advances in leaps and bounds, human nature stays much the same. There’s always somebody out to take advantage of the system ,and of you, while we remain always sympathetic to those we care about. The game hasn’t changed – just the tools of the game, and the ease with which the unscrupulous can run their scams on a wider swath of folks. And the tools are ever more easily and cheaply accessible. 

So, what to do?

There are simple steps to make it harder for unscrupulous folks to take advantage. First of all, educate yourself and your elder relations and loved ones about the risks. There’s no need to get paranoid about it – just to take some practical steps.

• DO enable two-factor authentication (2FA) , one of the most important, but simple security practices. 
• DON’T share passwords. (There are a few exceptions for the elderly or ill.)
• DON’T use dictionary words or favorite colors, cars, kids, or teams as your passwords.
• If passwords are hard to remember, DO use password managers, such as NordPass, Dashlane, 1Password, or Last Pass.
• DON’T give away your voice to be recorded by those you don’t know, no matter how much fun it looks like.
• DON’T click on links in emails or texts, especially from financial institutions. Rather, use your browser and type in the website of the institution. Most offer apps for direct access that are generally safe.
• DO look for odd misspellings or extra words in email addresses and domains. Website domains don’t have typos. For instance Welsfargo.com is not Wellsfargo.com (it’s missing an “l.”). Applepayments.com is not Apple.com/payments.
• DON’T open attachments to emails or texts if you’re not sure or don’t know who the sender is.

Modern tech and the Internet make so many things available for us and in many cases, easier for us. While there are always bad actors, and while AI tools make it easier for those bad actors to fool us, some sensible practices make it more difficult for those bad actors to fleece us of our hard-earned resources

Checking up on our elders and educating them about some of these simple steps can make them and you safer, help you both sleep at night, and even bring you a little bit closer, and after all, what could be better than that?

The post AI and Elder Abuse appeared first on Burgess Forensics.