xAI

Grok 4.5

• Released 14 July 2026
• 1 independent or government test
• 2 tests by the provider
• 2 source documents

Independent and government tests

• AI Security Leaderboard · FAR.AI, independent evaluator · Misuse & jailbreak

  for grok 4.5, the most vulnerable of the four, just $58 was enough to develop an attack that pulled harmful information across every domain we tested.

  FAR.AI · 29 July 2026 · Read the source

The provider's own tests

• Hackerbench · The provider · Cyber

  hackerbench is a spacexai internal ctf-style suite of cybersecurity and red-teaming tasks that require chaining skills (recon, exploitation primitives, and reasoning about vulnerable systems) in a constrained environment.

  xAI · 14 July 2026 · Read the source
• sycophancy · The provider · Deception & sabotage

  Reported result: sycophancy ↓ 0.01%

  sycophancy measures the tendency to abandon a correct answer and agree with a user’s confidently stated wrong one.

  xAI · 14 July 2026 · Read the source

No document we read states: a safety framework, whether a capability level was crossed, a safeguard. No incident on record names this version.

Some documents we hold are stored in lower case only, so their quotes are shown in lower case. The source link opens the original.

Compare with other versions