Gray Swan
An AI safety research organization focused on adversarial attacks against large language models and security evaluation. It operates red-teaming benchmarks including an indirect prompt injection arena.
Timeline (last 90 days)
Through built-in adversarial training and real-time safety checks, Gemini 4 Argon is in a leading position on Gray Swan's indirect prompt injection benchmark, reducing the attack success rate of 15 adaptive attempts to 0.7%.
Google says Argon leads on the Gray Swan indirect prompt injection benchmark, with an attack success rate of 0.7% across 15 adaptive attempts.
OpenAI's GPT-6 Astra system card reports a 99.99% defense rate against direct injection, but attackers still achieved an 8.5% success rate in Gray Swan's external indirect injection arena.
Argon claims to be the model currently most resistant to indirect prompt injection attacks, leading on Gray Swan's IPI benchmark.
All Facts (4)
Through built-in adversarial training and real-time safety checks, Gemini 4 Argon is in a leading position on Gray Swan's indirect prompt injection benchmark, reducing the attack success rate of 15 adaptive attempts to 0.7%.
70%UnverifiedGoogle says Argon leads on the Gray Swan indirect prompt injection benchmark, with an attack success rate of 0.7% across 15 adaptive attempts.
50%UnverifiedOpenAI's GPT-6 Astra system card reports a 99.99% defense rate against direct injection, but attackers still achieved an 8.5% success rate in Gray Swan's external indirect injection arena.
50%UnverifiedArgon claims to be the model currently most resistant to indirect prompt injection attacks, leading on Gray Swan's IPI benchmark.
50%