SecurityWeek’s weekly cybersecurity roundup highlights OpenAI banning a coordinated set of ChatGPT accounts linked to a Cambodia-based operation that used the model for investment, romance, gambling, and law enforcement impersonation scams. Amgen confirmed unauthorized access to data stored in third-party cloud environments in July 2026, with proprietary information and patient protected health information exfiltrated, though no impact on products, manufacturing, financial systems, or patient care was seen. Apple limited the number of vulnerability submissions researchers can have in its bug bounty program after a surge of low-quality, AI-hallucinated reports; cybersecurity firm Bynario hit the new cap after using ChatGPT to surface more than 50 macOS issues, including a privilege-escalation exploit it could not immediately report. The FCC is drafting rules to block imports of new Chinese optical transceivers used inside data centers, aiming to reduce risks of data theft, malware, or service disruption in AI infrastructure, with officials hoping to finalize the measure this year. A long-running supply chain compromise of the QuickFox VPN and game-accelerator app delivered a trojanized Electron installer that executed a JavaScript loader and installed the FDMTP implant on Windows systems; QuickFox removed the malicious components after Fortinet’s disclosure. Multiple models of Zbtlink and rebranded cellular routers come pre-loaded with an implant.