Your work emails may outlive the company that employed you.

Google Just Put a $10 Million Price Tag on Employees’ Old Emails

Your work emails may outlive the company that employed you.

Spirit Airlines is gone.

Its planes are being dealt with.

Its employees have been laid off.

Its operations have stopped.

But something remarkably valuable survived:

The conversations its employees left behind.

Google has agreed to pay $10 million in Spirit Airlines’ bankruptcy proceedings for access to a massive collection of the airline’s internal business data.

The reason?

Among other product-development uses:

Training artificial intelligence.

And the scale of what’s being sold is extraordinary.

Approximately:

100 million emails.

500 million Microsoft Teams messages.

Plus internal documents, spreadsheets, calendars, operational information, marketing data, productivity data and software.

Think about what that actually represents.

Years of employees:

Asking questions.

Solving problems.

Arguing.

Making decisions.

Explaining procedures.

Scheduling meetings.

Fixing mistakes.

Writing reports.

Collaborating.

Managing an airline.

Spirit’s aircraft were obviously valuable physical assets.

But in the AI economy, there was apparently another asset sitting quietly on its servers:

A gigantic recording of how thousands of humans actually work.

Why Would Google Want 500 Million Teams Messages?

Because AI companies have an enormous problem.

The internet contains staggering amounts of information.

But internet text isn’t necessarily a good representation of how employees actually perform their jobs.

Wikipedia can teach an AI about aviation.

A corporate archive can potentially teach it how people operate an airline.

Consider what exists inside years of internal communications.

Someone reports a problem.

Someone else diagnoses it.

A manager escalates it.

Employees debate possible solutions.

A decision gets made.

Someone implements it.

Something goes wrong.

They fix it.

Multiply that across millions of conversations.

You’re no longer looking at a collection of messages.

You’re looking at something resembling an enormous dataset of:

Problem → reasoning → decision → action → outcome.

That’s incredibly interesting training material for AI systems designed to perform knowledge work.

The Runner-Up Tells You Something Important

Google wasn’t alone.

Mercor reportedly bid $7.5 million for the dataset and is the backup purchaser if Google’s transaction doesn’t close.

Mercor operates in the AI ecosystem and works with human expertise and data used to improve AI systems.

That makes this more interesting than Google simply buying leftover corporate software.

There was competitive bidding for the information itself.

Corporate exhaust has become an asset class.

We’ve Seen a Tiny Version of This Before

There’s a fascinating precedent.

Enron.

When Enron collapsed, roughly half a million employee emails eventually became a famous public research dataset.

The Enron Email Dataset has subsequently been used for decades by researchers studying:

Natural-language processing.

Spam detection.

Social networks.

Organizational communication.

Email classification.

Machine learning.

It became extraordinarily valuable precisely because authentic corporate email is difficult to obtain.

Now compare roughly half a million Enron messages with:

100 million Spirit emails.

And then add:

500 million Teams messages.

The difference isn’t merely size.

Teams captures a different kind of communication.

Shorter.

Faster.

More conversational.

More informal.

More collaborative.

Many workplace conversations that would once have occurred verbally or disappeared entirely are now permanently recorded in Slack, Teams and similar platforms.

We created an extraordinarily detailed dataset of how modern offices function without necessarily realizing we were creating one.

The Employees Weren’t Writing AI Training Data

This is where the story becomes uncomfortable.

An employee writing:

Hey, did we ever figure out why that report keeps failing?

isn’t thinking:

“I’m contributing another sample to a future machine-learning corpus.”

They’re doing their job.

When employees communicated with coworkers, they understood they were using corporate systems.

They presumably understood the company retained those records.

But there’s a meaningful difference between:

“My employer stores my Teams messages.”

and:

“Years after I write this, these conversations might become an asset sold during bankruptcy to train someone else’s artificial intelligence.”

That’s the ethical question this case puts directly on the table.

The Data Is Supposed to Be De-Identified

There is an important safeguard.

Reporting says a third party will process the information before delivery to Google.

Personally identifiable information is supposed to be removed, and customer information isn’t part of the transaction.

That’s significant.

This isn’t Google simply receiving a searchable inbox containing employee names, customer records and credit-card information.

But de-identification doesn’t eliminate the larger question.

Who owns the knowledge created through everyday work?

The employee?

The employer?

The bankruptcy estate?

And if that information has economic value after the company dies, should employees have any say in how it’s subsequently used?

Legally, workplace communications created on company systems generally belong to the employer, subject to applicable contracts, policies and privacy laws.

AI is making the implications of that old reality much more visible.

Bankruptcy Changes How You Look at Data

Imagine a company shuts down.

What remains?

Buildings.

Computers.

Vehicles.

Furniture.

Patents.

Domain names.

Software.

Customer relationships.

Traditionally, those are the assets people expect to see sold.

Now add:

Every email employees ever wrote.

Every Teams conversation they ever had.

Every internal document they created.

Every workflow they developed.

Every operational problem they solved.

AI has potentially changed the liquidation value of information.

A database that once represented storage expense can now represent training material.

That’s a remarkable economic shift.

Your Company May Be Sitting on an AI Dataset Right Now

Forget Spirit for a moment.

Think about your own Microsoft 365 environment.

How many years of email exist?

How many Teams messages?

How many SharePoint documents?

How many support tickets?

How many meeting transcripts?

How many recorded calls?

How many internal procedures?

How many customer-service conversations?

How many Slack messages?

How many CRM notes?

Ten years ago, much of that was considered historical business data.

Today it can potentially be something else:

A dataset describing how your organization thinks.

That’s valuable.

And anything valuable needs governance.

This Creates a New Data-Protection Question

Most businesses ask:

How long do we need to retain this data?

AI gives us another question:

What could someone eventually do with it?

That’s much harder.

Your employee handbook may explain that corporate email belongs to the company.

But does your privacy policy explain whether employee communications can someday be:

Analyzed by AI?

Used to train models?

Licensed?

Sold?

Transferred during an acquisition?

Transferred during bankruptcy?

De-identified and monetized?

Most organizations wrote their data-retention policies before anyone seriously contemplated these possibilities.

They should revisit them.

“Deleted” and “Gone” Aren’t Always the Same Thing

Businesses should also understand where information actually exists.

An employee deletes an email.

Is it gone?

Maybe not.

It might still exist in:

Retention policies.

Litigation holds.

Backups.

Archives.

Security platforms.

Journaling systems.

Cloud repositories.

Third-party backup products.

eDiscovery systems.

The same applies to Teams and other collaboration platforms.

Modern businesses deliberately retain enormous amounts of information for legal, compliance and operational reasons.

That’s often necessary.

But retention has a security consequence:

You cannot lose data you no longer possess.

Every year of retained information increases the historical dataset that potentially exists during a breach, acquisition, lawsuit—or bankruptcy.

Don’t Retain Everything Forever Just Because You Can

This is where your MSP, cybersecurity team, attorneys and compliance professionals need to work together.

Data retention shouldn’t be:

“Storage is cheap, keep everything.”

Organizations should establish defensible retention schedules based on:

Legal requirements.

Regulatory obligations.

Operational needs.

Litigation requirements.

Contractual commitments.

Security risk.

Privacy.

Different information deserves different retention periods.

Keeping unnecessary information indefinitely creates unnecessary liability indefinitely.

There Is Also a Cybersecurity Gold Mine Here

Think about this dataset from an attacker’s perspective.

Corporate communications can reveal:

Internal terminology.

Organizational structure.

Vendor relationships.

Technology platforms.

Business processes.

Employee behavior.

Escalation procedures.

Historical incidents.

Internal projects.

Security discussions.

Even when obvious personal information is removed, organizational knowledge can remain extraordinarily valuable.

That’s why companies shouldn’t think about email security only as:

“Prevent someone from reading today’s inbox.”

A compromised Microsoft 365 environment may expose years of corporate memory.

Law Firms Should Be Extremely Careful

Imagine this principle applied to a law firm.

Years of internal email and Teams messages could contain:

Litigation strategy.

Client discussions.

Negotiation approaches.

Privileged information.

M&A activity.

Personnel matters.

Investigations.

Even where legal and ethical rules impose substantial restrictions on transferring or using such information, the underlying lesson remains:

Corporate communications can become enormously valuable datasets.

Law Firm IT needs retention, classification and access controls designed around that reality.

Healthcare Has an Even Higher Bar

Healthcare organizations face HIPAA and other privacy obligations that make sensitive patient information fundamentally different from ordinary corporate communications.

But they also generate massive quantities of operational data.

Internal workflows.

Scheduling communications.

Billing processes.

IT tickets.

Administrative conversations.

AI makes previously mundane operational information potentially valuable.

Healthcare IT teams therefore need clear data classification.

What is PHI?

What is employee information?

What is operational data?

Who owns it?

How long is it retained?

Who can use it?

Could it ever be provided to an AI system?

“It’s internal” is no longer a sufficient data classification.

Schools Should Think About This Too

Schools increasingly generate enormous digital archives.

Email.

Google Workspace.

Microsoft 365.

Student systems.

Staff chats.

Learning platforms.

Documents.

Recordings.

AI tools.

School Technology leaders need policies covering not merely storage and cybersecurity, but future use.

Especially when student information or employee communications are involved.

Employees Need to Understand One Brutal Rule

Don’t use company systems as if they’re personal systems.

Your corporate email account isn’t your diary.

Teams isn’t your private living room.

Slack isn’t disappearing conversation.

Your work laptop isn’t your personal computer.

That doesn’t mean employees should be paranoid.

It means they should understand the environment they’re communicating in.

If something is deeply personal and unrelated to work, don’t put it in the corporate archive unnecessarily.

Because corporate information can survive:

Your resignation.

Your termination.

Your manager.

The CEO.

An acquisition.

And apparently:

The company itself.

Businesses Need an AI Data Governance Policy Now

This shouldn’t wait until bankruptcy.

Every organization adopting AI should answer:

What corporate information may be submitted to AI systems?

Which AI vendors are approved?

Can vendors train on our information?

How long do they retain prompts?

Are employee communications included?

What happens to uploaded documents?

Can confidential information be used?

Can customer information be used?

Who approves new AI tools?

What happens when an employee leaves?

And critically:

What rights exist over our data if the vendor—or we—cease operations?

This is becoming part of cybersecurity and data protection.

Your MSP shouldn’t merely secure where your data lives.

Organizations increasingly need to understand where their data can go.

The $10 Million Lesson

Spirit’s wind-down began in May 2026 after years of financial problems.

But its digital history didn’t disappear when the airplanes stopped flying.

Google looked at that history and reportedly saw enough potential value to offer:

$10,000,000.

Not primarily for passenger profiles.

Not for a pile of old laptops.

For internal business information and software that could help develop products and train AI.

That’s the part every executive should understand.

Your organization’s emails, chats, documents and workflows aren’t merely records anymore.

Collectively, they may represent a model of how your company operates.

And models of how humans actually perform work are becoming extremely valuable in the AI economy.

So before you hit Send on the next Teams message, remember something uncomfortable:

The company may eventually disappear.

Your message might not.

70% of all cyber attacks target small businesses, I can help protect yours.

#Cybersecurity #ArtificialIntelligence #DataPrivacy #DataProtection #ManagedIT

Spirit Airlines died. Its employees’ emails didn’t. Google just bid $10 million for them.