Forgotten Accesses Expose Internal Data to AI Agents

Unrecorded SaaS connections, forgotten API keys, or temporary development spaces can allow AI agents to access all internal data. Few organizations would be aware of these accesses. The problem is not limited to tools used without authorization: rights that were initially approved can persist after a trial.
Tokens Can Survive Rights Revocation
Revoking a compromised key or a risky integration does not necessarily cut off all access pathways. Other agents or integrations may cache their own tokens, particularly those used for refreshing permissions. These elements can then automatically renew access in the background.
Credentials and delegated permissions concentrate most of the difficulties: OAuth consents, abandoned API keys, or service accounts. Another mechanism, known as "permission flattening," pertains to AI processing chains, particularly retrieval-augmented generation (RAG): they can ingest data using a service account with extensive rights.
Inadvertent accesses would pose a greater risk than deliberate transmissions to unapproved tools. In many cases, an integration authorized for a brief experiment retains its rights once the experiment is over.
An Employee Can Authorize a Connection Without Oversight
A simple click on "Allow" can trigger unauthorized access to company information. When employees can give their own consent, a connection can be validated without going through purchasing or the IT team. The tools in question may request extensive rights over messaging systems, shared directories, or customer relationship management software.
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisissez votre rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
Among common uses is an agent programmed by a sales team to automatically update customer records within the organization. Agents linking multiple SaaS applications are particularly difficult to detect and control, as their operation remains entirely in the background.
Trials Can Scatter Internal Data
On the development side, creating an API key to access customer data and forgetting an environment used to test an agent are common situations. Abandoned development spaces pose a persistent challenge. Cloud-hosted platforms facilitate copies to locations unknown to IT or security teams.
AI exacerbates this difficulty: an API key can directly link a third-party large language model or agent to a data repository. Projects can also leave behind indexes, test datasets, and logs outside existing governance frameworks.
The majority of security professionals surveyed do not see this as a new problem. Cloud and SaaS software had already raised comparable issues of dispersion and control two decades ago. What has changed is the speed and extent of the phenomenon; the pace of AI development and deployment would exceed the tracking capabilities of most teams responsible for cybersecurity and data governance.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.