Follow us on Google News
Google has agreed to strengthen privacy protections in its proposed $10 million purchase of business data belonging to the collapsed US carrier Spirit Airlines, following objections from employee unions concerned about sensitive information being used to train artificial intelligence systems.
The agreement, disclosed on Tuesday, introduces stricter requirements for removing identifying information from company records. It also excludes additional categories of employee data from the planned sale.
The development removes a major objection to the transaction ahead of a bankruptcy court hearing scheduled for November 17. However, the sale has not yet received final judicial approval.
Google Agrees To Additional Employee Data Protections
The Association of Flight Attendants-CWA, which represented Spirit Airlines flight attendants, said the agreement followed months of opposition from labour organisations and privacy advocates.
Under the revised terms, several categories of sensitive employee information will be excluded from the data transferred to Google.
These include medical and health information, drug and alcohol testing records, disciplinary documents, workplace safety programme records and identification documents.
The unions also secured stronger requirements for removing information that could identify individual employees.
According to the flight attendants’ association, the process will follow ISO 27559, an international standard addressing the de-identification of personal data.
An independent verification specialist, selected in consultation with the unions, will examine the process.
Google will also be prohibited from attempting to identify individuals again using the processed information.
The company must handle the data within a secure environment and treat it as confidential.
Following these commitments, the unions withdrew their objections to the proposed sale.
Why Google Wants Spirit Airlines’ Business Records
Spirit Airlines ceased operations in May 2026, leaving its remaining assets to be handled through bankruptcy proceedings.
As part of that process, Google won an auction in August to purchase a substantial collection of the airline’s internal business information.
The agreed purchase price was $10 million.
The dataset includes corporate emails, spreadsheets, calendars, workplace communications and information related to business operations.
Court records indicate that the proposed transaction covers approximately 100 million emails and 500 million workplace chat messages.
Other records concern marketing, productivity, financial operations and aircraft-related activities.
Google plans to use the material to improve its products and AI models.
Such business information can help AI systems understand how organisations manage complicated workflows, communicate internally and make operational decisions.
However, company records may contain references to employees, customers and other individuals, even when the dataset is primarily intended for business analysis.
Google has maintained that it does not intend to acquire identifiable personal information and that relevant records will undergo processing before the company receives them.
The new safeguards place additional requirements around that process.
What De-Identification Means For AI Training
De-identification refers to removing or changing information that could reveal a person’s identity.
For example, a company may remove names, email addresses, employee identification numbers or other direct identifiers from a database.
The aim is to make the remaining information useful for analysis without exposing the people associated with it.
However, removing names does not necessarily guarantee anonymity.
A dataset may still contain combinations of information, such as job titles, dates, locations or unusual events, that could allow someone to identify an individual.
This concern becomes particularly important when large collections of records are processed using advanced AI tools.
Privacy advocates have warned that information drawn from several sources can sometimes be combined to reveal identities that were not immediately obvious.
The Spirit Airlines dispute therefore raises questions extending beyond a single corporate transaction.
As companies increasingly use internal records to train AI systems, they must consider whether employees originally understood or agreed to such uses of their information.
The new agreement attempts to address these risks through stronger technical requirements, independent review and restrictions on how Google may use the data.
Union Objections Delayed The Proposed Sale
The dispute began after employee representatives raised concerns about the proposed transaction during Spirit Airlines’ bankruptcy proceedings.
The flight attendants’ union argued that existing protections did not adequately address the risks posed by transferring workplace information for AI development.
Other labour organisations subsequently joined the objections.
The dispute also attracted attention from US lawmakers.
In October, a group of 121 members of Congress urged stronger safeguards, warning that removing obvious identifiers might not be sufficient to protect employee privacy.
Court proceedings concerning the sale were delayed while these concerns were examined.
The latest agreement resolves the unions’ immediate objections, although the court must still determine whether to authorise the transaction.
For businesses developing AI systems, the case highlights how employee privacy can become a major issue when corporate data is sold during restructuring or liquidation.
It also demonstrates that information collected for ordinary workplace purposes may require additional safeguards before being transferred for an entirely different commercial use.
What This Means For You
Employees should review workplace privacy policies to understand how their information may be retained, transferred or used after leaving an organisation. Businesses planning to use employee records for AI development should assess privacy risks and establish safeguards before transferring the data.
