CYBERSECURITYTRACKER
TRACKING6,626 stories in this site build1,366 vulnerability news stories in this site build
Permanent story citation

Anthropic’s Opus 5 Is Better at Resisting Prompt Injection

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 3655

As cited

Copy frozen at (site build).

ai security

Anthropic’s Opus 5 Is Better at Resisting Prompt Injection

Anthropic's Claude Opus 5 demonstrates improved resistance to prompt injection attacks, reducing successful attack rates to 2.0% within 15 attempts compared to 5.5% for its predecessor Opus 4. The model significantly outperforms competing offerings, with the most robust non-Claude alternative succeeding at over eight times the failure rate of Opus 5.

Why it matters: Security teams evaluating large language models for sensitive applications should consider prompt injection resilience when selecting AI tools, as Opus 5's improved defenses reduce a key attack surface for jailbreaking and malicious prompt attacks.

Source published
First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

ai security

Anthropic’s Opus 5 Is Better at Resisting Prompt Injection

Anthropic's Opus 5 model shows improved resistance to prompt injection attacks compared to its predecessor Opus 4.8 and other large language models, reducing attacker success rates from 5.5% to 2.0% on the IPI benchmark within 15 attempts. Opus 5 outperformed competing models including GPT 5.6 variants, which showed success rates ranging from 20% to 43.9% in the same test. The results indicate ongoing progress in mitigating prompt injection vulnerabilities in specific scenarios, though the source notes complete prevention remains theoretically impossible.

Why it matters: Security teams evaluating large language models for sensitive applications should consider these resilience metrics when comparing Claude and competing platforms for deployment.

Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary