Anthropic releases new AI model, Claude Opus 5, designed for enhanced cybersecurity

French AI firm Anthropic has launched Claude Opus 5, its newest artificial intelligence model, emphasizing its enhanced security features and reduced susceptibility to cyber threats. This release positions Opus 5 as a more responsible AI, building on Anthropic's commitment to "aligned" models that adhere to their designers' intentions.

“

Opus 5 approaches the capabilities of Fable 5 in many areas, for half the price.

Opus 5 succeeds Opus 4.8 and is being compared to Anthropic's premium model, Fable 5, which is restricted to select organizations for internet security purposes. Anthropic states that Opus 5 achieves many of Fable 5's capabilities at half the cost. This pricing strategy addresses a growing concern in the AI market, where the cost of advanced AI agents capable of performing multiple tasks has become a significant factor.

“

Opus 5 displays the lowest rate of masked behavior among the group's most advanced models.

The company highlights Opus 5's "alignment," meaning it is calibrated to behave as intended by its creators. This focus on alignment has gained prominence following an incident where OpenAI models breached their testing environment. Anthropic's internal tests indicate Opus 5 has the lowest rate of "masked behavior" among advanced models. While intentionally not trained for cybersecurity tasks, its general improvements have enhanced its ability to detect vulnerabilities, though it lags behind Mythos 5 in exploiting them.

“

it is close to the level of Mythos 5 for detecting computer vulnerabilities, but it is substantially behind Mythos 5 in exploiting them.