CVE-2026-57474: Deloitte AI Assist Information Disclosure

CVE-2026-57474 Overview

CVE-2026-57474 affects Deloitte AI Assist for Customer, a customer-focused AI assistance service. Public-facing API endpoints accepted unauthenticated requests and returned configuration information to any caller. The exposed data reduced the reconnaissance effort required by attackers targeting the service. Deloitte remediated the issue on 2026-03-25 by restricting network access and enforcing authentication on the affected endpoints. The weakness is tracked as CWE-200: Exposure of Sensitive Information to an Unauthorized Actor.

Critical Impact
Unauthenticated attackers could retrieve configuration information from public-facing API endpoints, aiding reconnaissance for subsequent attacks against Deloitte AI Assist for Customer.

Affected Products

Deloitte AI Assist for Customer (service, all versions prior to remediation)
Public-facing API endpoints exposed by the service
Deployments accessible over the network prior to 2026-03-25

Discovery Timeline

2026-03-25 - Deloitte restricted network access and enforced authentication on the exposed endpoints
2026-07-10 - CVE-2026-57474 published to NVD
2026-07-16 - Last updated in NVD database

Technical Details for CVE-2026-57474

Vulnerability Analysis

The vulnerability is a configuration information disclosure issue in Deloitte AI Assist for Customer. Several API endpoints were reachable over the network without any authentication requirement. When queried, these endpoints returned configuration details about the service.

An unauthenticated remote attacker could send HTTP requests to the exposed endpoints and retrieve responses containing configuration data. This data does not provide direct access to customer records or code execution, but it lowers the effort required to plan further attacks. The disclosed information supports enumeration of internal components, service versions, and integration parameters.

The attack requires no user interaction and no privileges, and it is exploitable across the network. The impact is limited to confidentiality of configuration data, with no direct effect on integrity or availability of the service.

Root Cause

The root cause is a missing access control on public-facing API endpoints. Endpoints that should have required authentication and network-level restrictions were reachable by anonymous clients. The service returned configuration data in response to these unauthenticated requests, mapping directly to [CWE-200].

Attack Vector

Exploitation occurs over the network against the exposed HTTP API. An attacker enumerates the public endpoints of AI Assist for Customer and issues unauthenticated requests. The service responds with configuration information that supports reconnaissance, such as identifying downstream services, endpoint paths, and configuration parameters that can inform targeted follow-up requests.

No verified proof-of-concept code is published for this issue. See the ZeroTolerance AI Assist Advisory and the CISA CSAF Report for additional technical detail.

Detection Methods for CVE-2026-57474

Indicators of Compromise

Anonymous HTTP requests to Deloitte AI Assist for Customer API endpoints prior to 2026-03-25
Access log entries containing successful 200 OK responses to unauthenticated requests against configuration-related paths
Repeated enumeration patterns from a single source IP targeting the AI Assist API surface

Detection Strategies

Review historical web access logs for requests to AI Assist for Customer endpoints that returned data without an Authorization header or session cookie
Correlate unauthenticated API responses with subsequent authenticated activity from the same source addresses
Alert on high-volume enumeration behavior against public API paths and JSON responses containing configuration keys

Monitoring Recommendations

Continuously monitor perimeter API gateways for unauthenticated 2xx responses on endpoints that should require authentication
Track configuration disclosure attempts as reconnaissance events in the SIEM and pivot to source-IP investigation
Establish baselines for API request volume per endpoint to detect scanning activity

How to Mitigate CVE-2026-57474

Immediate Actions Required

Confirm that all AI Assist for Customer API endpoints reject unauthenticated requests
Restrict network access to management and configuration endpoints to trusted networks only
Review historical logs for any anonymous access to the previously exposed endpoints and assess exposure of specific configuration data
Rotate any secrets, tokens, or integration credentials that may have appeared in the disclosed configuration responses

Patch Information

Deloitte remediated the vulnerability on 2026-03-25 by restricting network access and enforcing authentication on the previously exposed endpoints. The fix is applied on the vendor-hosted service, so no customer-side patch is required. Refer to the CVE Record for CVE-2026-57474 and the ZeroTolerance Deloitte Advisory for advisory details.

Workarounds

Place API endpoints behind an authenticated reverse proxy or API gateway that enforces token validation
Apply IP allow-listing to limit endpoint reachability to known client networks
Remove verbose configuration data from public API responses and return only the minimum fields required

bash

# Example: enforce authentication on an API gateway route (NGINX)
location /api/aiassist/ {
    auth_request /_auth;
    allow 10.0.0.0/8;
    deny all;
    proxy_pass https://backend.internal/aiassist/;
}