Tag: machine learning risks
3 articles
OpenAI's GPT-6.1 Terra Exposes Destructive-Action Flaws Despite Safety Gains
While OpenAI's GPT-6.1 Terra shows promising safety gains, with 42% fewer severe flags than its predecessors, it surprisingly falls short in specific areas, revealing flaws that could lead to destructive actions. This nuanced picture highlights the ongoing challenge of balancing safety and performance in AI development.
GuardFall Exposes AI Coding Agents to Shell Injection Risks
Researchers at Adversa AI have uncovered a shocking weakness, dubbed GuardFall, that lets advanced open-source coding agents slip past safety filters and execute destructive shell commands, exposing them to shell injection risks. This gap between text-based checks and shell execution leaves a trail of vulnerability wide open to exploitation.
Enterprises Lag in Securing Autonomous AI Agents
Most organizations are struggling to keep pace with the rapid evolution of autonomous AI agents, which can introduce new risks and behaviors at machine speed. As these agents increasingly handle sensitive data, enterprises face a pressing need to update their security strategies and tools to mitigate the emerging threats of shadow AI and over-permissioned agents.