What: Cybercriminals are using OpenAI’s API as a covert command channel to deploy persistent backdoors and evade detection in retail digital environments.

Why it is important: The abuse of trusted AI APIs for cyberattacks exposes new vulnerabilities in retail IT, requiring urgent updates to security and vendor oversight.

The discovery of the SesameOp backdoor, which utilises OpenAI’s API as a stealthy command-and-control channel, marks a significant escalation in the sophistication of cyber threats targeting retail digital environments. By exploiting legitimate AI APIs, attackers can maintain persistent, covert access to compromised systems, bypassing traditional security measures and blending in with normal network activity. This approach leverages trusted developer tools and advanced obfuscation techniques, making detection and remediation particularly challenging for retail IT teams. The campaign’s focus on long-term espionage and operational control raises serious concerns about data security, customer privacy, and the integrity of retail operations. As AI-powered integrations and third-party APIs become increasingly embedded in retail workflows, the sector faces heightened risks from both external and supply chain attacks. These developments underscore the urgent need for retailers to strengthen cybersecurity governance, enhance vendor oversight, and implement continuous monitoring of all AI and API-driven processes to protect against emerging threats and safeguard business continuity.

IADS Notes: The Robin Report (August 2025) and BCG (August 2025) both highlight how AI systems and APIs introduce new vulnerabilities, with 41% of breaches linked to third-party providers and average losses exceeding £1.4 million. RH-ISAC (April 2025) and Retail Week (August 2025) document the rise of sophisticated, AI-driven cyberattacks on retail, while Trustwave (May 2025) details the evolving tactics of advanced cybercrime groups exploiting both traditional and AI-powered vectors.

Microsoft detects "SesameOp" backdoor using OpenAI's API as a stealth command channel