惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Microsoft Security Blog
Microsoft Security Blog
博客园 - 聂微东
aimingoo的专栏
aimingoo的专栏
J
Java Code Geeks
腾讯CDC
大猫的无限游戏
大猫的无限游戏
H
Hackread – Cybersecurity News, Data Breaches, AI and More
Google DeepMind News
Google DeepMind News
博客园_首页
F
Fortinet All Blogs
小众软件
小众软件
Apple Machine Learning Research
Apple Machine Learning Research
H
Help Net Security
博客园 - 【当耐特】
量子位
博客园 - 叶小钗
M
MIT News - Artificial intelligence
酷 壳 – CoolShell
酷 壳 – CoolShell
月光博客
月光博客
MyScale Blog
MyScale Blog
爱范儿
爱范儿
The Cloudflare Blog
N
Netflix TechBlog - Medium
T
Tailwind CSS Blog

Help Net Security

ChatGPT advanced account security adds passkeys and hardware keys Week in review: High-severity LPE vulnerability in the Linux kernel, cPanel 0-day exploited for months Automating Pentest Delivery: A Step-by-Step Guide - PlexTrac Open-source privacy proxy masks PII before prompts reach external AI services Shadow AI risks deepen as 31% of users get no employer training Identity is the control plane for distributed infrastructure AI traffic is getting bigger, louder, and less predictable New infosec products of the month: April 2026 cPanel zero-day exploited for months before patch release (CVE-2026-41940) Cisco releases open-source toolkit for verifying AI model lineage Met Police face criticism for using AI to spy on their own officers Nine-year-old Linux kernel flaw enables reliable local privilege escalation (CVE-2026-31431) Hacker with a special interest in breaching sports institutions ends behind bars - Help Net Security IP Fabric MCP server adds governance and control to enterprise AIOps workflows - Help Net Security Aqua Compass MCP server enables real-time investigation and containment of runtime threats - Help Net Security Google brings instant email verification to Android, no OTP needed - Help Net Security If cyber espionage via HDMI worries you, NCSC built a device to stop it - Help Net Security Apple fixes iPhone bug that let FBI retrieve deleted Signal messages(CVE-2026-28950) - Help Net Security GopherWhisper APT group hides command and control traffic in Slack and Discord - Help Net Security OpenAI tackles a bad habit people have when interacting with AI - Help Net Security A year in, Zoom's CISO reflects on balancing security and business - Help Net Security Scenario: Open-source framework for automated AI app red-teaming - Help Net Security GDPR works, but only where someone enforces it - Help Net Security Ransomware, fraud, and lawsuits drive cyber insurance claims to new peaks - Help Net Security Google’s Workspace Intelligence promises privacy while running on your data - Help Net Security Cyberattack on French government agency triggers phishing alert - Help Net Security Claude Mythos finds 271 Firefox flaws, Mozilla believes zero-days are numbered - Help Net Security Prove Identity Platform connects verification, authentication, and fraud prevention - Help Net Security New Mirai variants target routers and DVRs in parallel campaigns - Help Net Security Acronis GenAI Protection gives MSPs control over AI usage and data risks - Help Net Security
Anthropic launches Claude Opus 4.8, prepares Mythos-class...
Sinisa Markovic · 2026-05-29 · via Help Net Security

Anthropic has released Claude Opus 4.8 and outlined plans for broader access to its Mythos-class models, which the company expects to make available to all customers in the coming weeks.

Claude Opus 4.8

Claude Opus 4.8 (Source: Anthropic)

Claude Opus 4.8 is available to all users, with pricing unchanged from Opus 4.7.

Anthropic highlighted improvements in model honesty, noting that Opus 4.8 is more likely to acknowledge when it lacks sufficient information and less likely to make unsupported claims.

“Early testers have found Claude Opus 4.8 to be more reliable and sharper in its judgement when it’s performing agentic tasks,” the company wrote.

According to Anthropic, evaluations showed Opus 4.8 was around four times less likely than its predecessor to allow flaws in code it generated to go unremarked.

Anthropic reported lower rates of deceptive behavior and misalignment in testing. Opus 4.8 performed at a level similar to Claude Mythos Preview on measures of whether systems act in accordance with user interests and instructions.

Alongside the new model, Anthropic introduced Dynamic Workflows, a feature available in research preview that allows Claude to take on larger tasks in Claude Code. The feature enables Claude to plan work, run hundreds of parallel subagents within a single session, and verify outputs before returning results to the user.

“For example, Claude Code with Opus 4.8 can now carry out codebase-scale migrations across hundreds of thousands of lines of code from kickoff to merge, with the existing test suite as its bar,” Anthropic noted.

Dynamic Workflows is available for Claude Code users on Enterprise, Team, and Max plans.

The release also includes controls that allow users to adjust how much effort the model spends on a task. The setting can be used to balance response speed, reasoning depth, and cost.

Anthropic also revealed plans for a new class of AI model that it says will surpass Opus in intelligence. Mythos-class models, currently being tested by a limited number of organizations through Project Glasswing, are expected to become available to all customers in the coming weeks once additional cyber safeguards are in place.

“There’s still more to be done: we’re working on developing and releasing models that provide many of the same capabilities as Opus at a lower cost,” the company concluded.