Anthropic is rolling out a public version of its Mythos AI model, but with guardrails barring its use in ​risky areas such as cybersecurity, after an earlier preview this year sent shockwaves globally for its ability to find software flaws. The ne…