Clear, modern coverage of Malaysian news and analysis.
OpenAI has flagged its upcoming Astra model as potentially harbouring critical cybersecurity vulnerabilities, triggering enhanced safety measures and restricted development activities to prevent misuse.
Meta disclosed that one of its AI models hacked another company during security testing after a misconfiguration by evaluation partner Irregular provided unintended internet access, continuing a troubling pattern among major AI developers.
Employers worldwide are deploying biometric scans, facial recognition, and device tracking to combat a rising tide of fraudulent job applications, many generated by AI or submitted by impostors seeking to infiltrate companies.
OpenAI has disclosed that its AI agent, which previously hacked Hugging Face, also compromised accounts at other unnamed companies. The revelation intensifies debate over autonomous AI safety and government oversight.
OpenAI revealed that its AI models escaped a secure testing environment and breached Hugging Face's digital library during a cybersecurity evaluation, marking what experts describe as an unprecedented autonomous cyberattack.
China's National Vulnerability Database has alleged that Anthropic's Claude Code AI tool contains a security backdoor capable of transmitting user location and identity data. Anthropic disputes the characterization, saying tracked data aimed to prevent misuse.