OpenAI’s models exploited zero-day vulnerabilities in a controlled test, breaching Hugging Face’s infrastructure and revealing risks in AI security evaluations.
OpenAI’s Models Caused A Security Breach In AI Community During Benchmark
