CVE-2026-12203: HKUDS AI-Trader Research Export Information Disclosure

Technical summary

CVE-2026-12203 is an information disclosure vulnerability in HKUDS AI-Trader affecting the Research Export component, specifically the /api/research/agents.csv endpoint. The flaw is classified as improper access control (CWE-284) combined with information exposure (CWE-200). The vulnerability allows remote, unauthenticated attackers to retrieve sensitive data without authentication or user interaction. Attack complexity is low, indicating no special conditions are required. The vendor's remediation enforces authentication and validates that requesters possess the research_exports capability before permitting access to export endpoints. The patch commit is 91a31aac1b0f4dbc6b8bef9f6eff0b7912e0bc65.

Business impact

Information disclosure through unprotected research export endpoints poses moderate risk to organizations deploying HKUDS AI-Trader. Attackers could harvest sensitive research data, agent configurations, or related metadata that may inform further attacks or competitive intelligence gathering. Given the public availability of exploit code, risk of opportunistic exploitation is elevated. The impact is primarily confidentiality-focused; integrity and availability are not affected. Organizations relying on this platform for sensitive research workflows should prioritize remediation.

Affected systems

HKUDS AI-Trader up to commit 74caf996f78dcc0c657df8365c8544678a16e215 is vulnerable. The product uses a rolling release model without traditional version numbers, so affected releases are identified by commit hash. The Research Export component is the specific attack surface. Organizations running any build prior to the patch commit should be considered at risk.

Exploitability

Exploitability is high. The vulnerability is network-accessible, requires no authentication, no user interaction, and has low attack complexity. Proof-of-concept code is publicly available, lowering the barrier for opportunistic threat actors. An attacker needs only network access and knowledge of the endpoint path to attempt data extraction.

Remediation

Apply the vendor patch (commit 91a31aac1b0f4dbc6b8bef9f6eff0b7912e0bc65) which implements authentication requirements and capability-based authorization for research export endpoints. Verify the patch has been deployed by confirming that unauthenticated requests to /api/research/agents.csv are rejected with authentication errors. No workarounds are documented; patching is the recommended action.

Patch guidance

Update HKUDS AI-Trader to a build incorporating commit 91a31aac1b0f4dbc6b8bef9f6eff0b7912e0bc65 or later. Because this product follows rolling releases, consult the vendor's release notes or advisory to identify the next available build that includes this patch. Test the update in a non-production environment first to ensure the authentication changes do not disrupt legitimate research export workflows. Verify that the research_exports capability is properly assigned to intended users and service accounts.

Detection guidance

Monitor for suspicious access attempts to /api/research/agents.csv from external or untrusted sources, particularly unauthenticated requests. Log analysis should flag any successful data exports from this endpoint prior to the patch date. Network-based detection can identify requests to this URI without proper authentication headers. Check application logs for repeated failed authentication attempts or permission denied responses, which may indicate reconnaissance or exploitation attempts.

Why prioritize this

While the CVSS score of 5.3 (Medium) reflects the lack of impact to integrity and availability, the presence of public exploit code and the ease of exploitation warrant prompt attention. The vulnerability enables direct data theft with minimal attacker effort. Organizations should prioritize patching within their standard security update cycles, treating this as a medium-urgency remediation task.

Risk score, explained

The CVSS 3.1 score of 5.3 reflects: network-accessible attack vector (no local access required), low attack complexity (no special conditions), no privilege requirement, no user interaction needed, and limited scope. The impact is restricted to confidentiality loss; integrity and availability are unaffected. The presence of public exploits elevates practical risk beyond the base score, warranting timely mitigation despite the moderate rating.

Frequently asked questions