Eighteen thousand five hundred Strava profiles identified from their morning runs. Le Monde identified 18,599 profiles traceable to roughly a hundred French military bases, probably linked to military personnel, contractors, or former Defence Ministry staff, using only the application’s public data. After one year of monitoring, 1,372 of the 18,599 profiles had been switched to private, approximately 7%.

The essentials

  • Wellness data can become exploitable for intelligence purposes when collected, cross-referenced, and analyzed in context: neither hacking nor malicious intent required.
  • Le Monde identified 18,599 Strava profiles, probably linked to military personnel, contractors, or former Defence Ministry staff, spread across roughly a hundred bases (StravaLeaks investigation, published in 2024 and 2026).
  • The same mechanism had exposed American bases as early as 2018: eight years after the 2018 American restriction in operational zones, geolocation risks persist.
  • Individual vigilance is insufficient: only 1,372 of the 18,599 profiles in question, approximately 7%, switched to private mode after one year of monitoring.
  • Durable protection requires private settings imposed by default on platforms and collective rules within exposed organizations.

Running reveals more than one might think

Strava is a sports-tracking application. It records your routes, your pace, your heart rate data. Millions of people use it to improve, share their performances, motivate themselves. Nothing in that description suggests a national security risk.

But here is what an application of this type does by design: it associates an identity with a location, and it does so repeatedly, at regular hours, over months or years. A soldier who goes running at 6:30 a.m. from Creil base three times a week does not disclose a secret by publishing one route. He discloses it by publishing two hundred successive routes.

Le Monde’s StravaLeaks investigation, whose episode on the 18,599 profiles was published on March 29, 2026, made this mechanism visible. Journalists applied a simple method: cross-referencing publicly geolocated data with the known footprints of French military installations. Result: 18,599 individual profiles traceable to roughly a hundred bases, from Camp de la Lande in Lorient to the offices of the Directorate-General for External Security in Paris. No system was hacked. No classified data was stolen.

The application simply functioned as designed.

The 2018 American precedent stopped nothing

The phenomenon had been documented as early as January 2018. Strava published a worldwide heatmap of its users’ activities on November 1, 2017. Australian researchers noticed that certain desert zones glowed with unusual brightness: the precise outline of American military bases in Syria, Afghanistan, and Africa appeared highlighted, traced by aggregated soldiers’ sports activities on the map.

The Pentagon banned the use of geolocation functions by its personnel in operational zones on August 3, 2018. But the ban targeted the extreme case, not the underlying problem. It applied to the soldier deployed in Kandahar, not to the intelligence technician jogging around Creil base.

Roughly eight years separate the signal of January 2018 from the episode published on March 29, 2026, but Le Monde’s follow-up investigation began in February 2024. In the interval, the number of Strava users increased significantly, smartwatches became widespread in militaries, and the mass of publicly available geolocated data grew accordingly. Geolocation risks persist on a large scale.

7%: the ceiling of individual vigilance

This mechanism of inertia is reinforced by the fragmentation of perceived responsibility: each individual assumes that other members of his or her organization have taken the step, which further reduces the probability that he or she will do it themselves. The larger the exposed group, the more this dilution effect is accentuated. Collective protection therefore requires a modification of the framework in which these individual decisions are made.

The most revealing data point from the StravaLeaks investigation concerns what happened after the profiles were identified.

After one year of monitoring, 1,372 of the 18,599 profiles had been switched to private, approximately 7%. Roughly nine out of ten profiles had not been made private.

This figure invalidates a common assumption: that information alone is sufficient to produce behavior change. The military personnel in question are not naïve: many have received operational security training. But the psychology of everyday life works differently from the psychology of alert. A sports application belongs to the register of leisure, wellness, sociability among colleagues. Reconfiguring privacy settings requires an active decision, an effort to step outside the habitual frame.

This inertia is a documented behavioral fact in many contexts, from retirement regimes to organ donation: default options play a determining role in mass behavior. Applied to data security, this principle has a direct consequence: entrusting protection to individual initiative amounts to accepting a structural rate of exposure.

The question then facing military and security institutions is not how to better train individuals. It concerns who sets the default parameters, and to what end.

Available measures for platforms

Strava offers privacy options. Users can make their activities private, hide their starting and ending addresses, disable their appearance in local rankings. These settings have existed for years.

But they are optional. By default, upon registration, the profile and activities are visible to “Everyone” and location data are aggregated into the global statistics that Strava publishes, though other initial settings are more restrictive. This architecture is not an oversight: it is a design choice that favors user engagement. The global heatmap, rankings between friends, community challenges: all these engagement mechanisms rest on data visibility.

The economic model and the security model are therefore in direct tension. Strava announced, after the 2018 episode, improvements in handling sensitive data. “Privacy zones” allow automatic masking of data near user-designated addresses. But this feature still requires active action: the user must configure their zones. It does not protect organizations whose members have not taken this step.

A different solution would be to offer, for certain categories of users or in certain geographic contexts, private settings by default. The user who wishes to share their data would do so by explicit choice, not by inertia. This architecture, called “privacy by design,” is encouraged by the General Data Protection Regulation, but it is not imposed as a standard for wellness applications. The European Union has laid regulatory foundations for personal data protection; their application to national security cases and mass-market platforms remains uneven.

This type of tension between civilian uses and security implications extends beyond sports alone. Comparable logic can be found in other domains where foreign investments and interests penetrate normally partitioned spaces, such as sovereign wealth funds establishing themselves in sports leagues, bringing with them access and data that no one had anticipated.

Obligations that organizations can impose

This third path runs into a limit specific to hierarchical institutions: operational security rules are designed for service time, not for out-of-frame behaviors. Extending their scope to personal uses requires legitimacy that organizations have not always secured with their personnel, which undermines buy-in and favors informal workarounds rather than durable practice changes.

Responsibility does not rest solely with platforms. Institutions that employ sensitive personnel have levers of their own.

The first path is regulatory: forbid or restrict the use of sports geolocation applications on certain sites or for certain functions. Some militaries have done this partially. But a total ban runs into practical problems: smartwatches are now workplace health tools in several military programs. Heart rate and recovery data serve medical monitoring of personnel. Eliminating the tool to eliminate the risk would eliminate its benefits as well.

The second path is technical: require members in question to maintain a fully private profile, monitored periodically. This measure already applies in certain special units. Its systematic deployment faces implementation obstacles and personnel resistance who perceive these restrictions as intrusions into their personal life outside work.

The third path, less explored, is collective: treat the management of sensitive personnel’s personal data as a question of institutional hygiene, the same as physical security procedures. This requires training that starts from the mechanism—how innocuous data becomes intelligence data through aggregation—rather than from a list of prohibitions. And it requires organizations to accept exercising influence over behaviors that occur outside work time, raising legal and institutional culture questions that French militaries have not yet resolved.

The next application already exists

Strava is not an isolated case. It is one example among a rapidly expanding category: applications that collect health, location, and behavior data with increasing granularity, store them on third-party servers, and render part of it public by design.

Current smartwatches record sleep, heart rate variability, physiological stress levels. Meditation applications track practice hours and locations. Nutrition platforms geolocate meal times. Individually, each of these data points is innocuous. Combined on the same person over a year, they permit detailed mapping of a person’s routines and associations.

The proliferation of wellness sensors increases structural risk over the medium term. Nations that have invested in open data collection and analysis capabilities, known as “OSINT” for open-source intelligence, possess means to rapidly aggregate these flows. The work performed manually by Le Monde’s journalists can be automated and accelerated by algorithmic processing.

The question for military and security institutions is therefore not whether this type of exposure will recur, but how quickly potential adversaries exploit it. The StravaLeaks investigation has made visible a mechanism known to specialists but absent from public debate. It identifies 18,599 profiles across roughly a hundred bases, probably linked to military personnel, contractors, or former staff, and records 1,372 profiles made private after one year of monitoring, approximately 7%.

These figures define the scope of a governance problem, not a one-off scandal. The responses that follow—default parameters imposed on platforms, collective rules within exposed organizations, training on mechanisms rather than symptoms—are known. The open question is that of political will to impose them on private actors whose model rests precisely on data visibility.


Sources

  1. Le Monde, StravaLeaks investigation, “How Le Monde located 18,000 French military personnel with a fitness app,” March 2026
  2. New York Times, “Strava’s Fitness Heatmap and the End of Secrets,” January 2018 (documented precedent of American bases)
  3. General Data Protection Regulation (GDPR), Regulation (EU) 2016/679, article 25, Privacy by design
  4. Journal d’un Progressiste, “Sovereign Wealth Funds Expose Sports Leagues to Their National Priorities”