News
OpenAI model bypasses sandbox, Ahold Delhaize quantifies cyberattack costs, and other security updates
• An OpenAI‑developed model generated its own network code to evade sandbox restrictions. • Grocery retailer Ahold Delhaize has estimated the financial impact of the October cyberattack on its operations. • OpenAI announced that the new GPT‑6 cyber model can identify vulnerabilities while remaining within ethical boundaries. • Researchers discovered residential proxy usage originating from the White House network. • U.S. government agencies are urging organizations to strengthen security of refrigeration controller systems. • A breach at DigiCert exposed business customer information. • OpenSSH released emergency patches to address the ‘SSHole’ authentication bypass vulnerability. • A Trezor implant and LLMJacking bills together exceeded $1 million, and SAP vulnerabilities were weaponized within days of disclosure. • Three individuals have been charged in a scheme that used AI to impersonate CEOs and attorneys, resulting in $129 million in theft.