Latest breakthroughs in AI and technology today

# Anthropic's Claude AI: A Whirlwind Month of New Products, Fixes, and Cybersecurity Breakthroughs **April 27, 2026** --- Anthropic, the AI safety company behind the Claude family of models, has had one of its busiest months yet — rolling out new creative tools, addressing a user backlash over performance issues, and unveiling a model with striking cybersecurity capabilities. ## Claude Design: From Ideas to Visuals in Seconds On April 17, Anthropic launched **Claude Design**, a new experimental product aimed at helping non-designers — such as founders, product managers, and marketers — turn ideas into polished visuals. Users can describe what they want in natural language, and Claude generates prototypes, slide decks, one-pagers, and other visual assets that can then be refined through direct edits or follow-up requests. Anthropic has positioned the tool as a complement to existing design platforms rather than a competitor. Outputs can be exported as PDFs, PowerPoint files, or URLs, and can even be sent directly to Canva for collaborative editing. Teams can also apply their own design systems to maintain brand consistency across projects. ## Claude Code Controversy: Anthropic Admits Engineering Missteps Not all the news has been positive. For much of April, users of **Claude Code** — the company's command-line coding assistant — reported a noticeable decline in output quality. After initially attributing the problems to user behavior and intentional optimization changes, Anthropic published a detailed engineering post on April 24 acknowledging that three separate internal missteps were to blame. The issues included a reduction in default reasoning effort, a caching bug that dropped thinking history from context, and a verbosity prompt change that degraded coding quality. All three have since been resolved, and Anthropic reset usage limits for all subscribers as a goodwill gesture. However, some users have expressed frustration with the company's handling of the situation, saying it took too long to acknowledge the root cause. ## Persistent Memory Comes to Managed Agents On April 23, Anthropic announced the public beta of **persistent memory for Claude Managed Agents** on the Claude Platform. The feature allows developers to build AI agents that retain and apply lessons across sessions, with memories stored as files on a filesystem that can be exported, edited, and managed through the API or the Claude Console. Early adopters include Netflix, Rakuten, Wisedocs, and Ando. Rakuten reported a 97% reduction in first-pass errors, a 27% cost reduction, and a 34% improvement in latency. The feature is workspace-scoped and observable, giving organizations control over what their agents learn and retain. ## Claude Mythos Preview and Project Glasswing Perhaps the most dramatic announcement of the month came from Anthropic's security research team. In early April, the company revealed **Claude Mythos Preview**, a new general-purpose model with exceptional capabilities in computer security tasks. The model was able to autonomously discover previously unknown (zero-day) vulnerabilities in real-world software — a capability that Anthropic described as a watershed moment for the cybersecurity industry. In response, the company launched **Project Glasswing**, a coordinated initiative to use Mythos Preview to help secure critical software infrastructure. The model is initially being made available to a limited group of industry partners and open-source developers so that defenders can begin reinforcing key systems before models with similar capabilities become widely accessible. ## Massive Compute Expansion and Valuation Surge Anthropic has also been expanding its infrastructure. On April 20, the company announced an expanded collaboration with Amazon for up to **5 gigawatts of new compute capacity**, and on April 6, it revealed a partnership with Google and Broadcom for additional gigawatts of next-generation computing power. Meanwhile, Bloomberg reported that venture capital firms have been offering Anthropic a preemptive funding round that would value the company at **$800 billion or more** — a figure that would rival or even surpass OpenAI. Anthropic has reportedly declined these offers so far. ## Looking Ahead With new product launches, massive infrastructure deals, and groundbreaking security research, Anthropic is clearly positioning Claude as more than just a chatbot. As the company navigates both rapid growth and the growing pains that come with it, the AI industry will be watching closely to see how it balances innovation with the transparency and safety principles it has built its brand upon. --- *This article was generated based on publicly available reporting and official announcements from Anthropic as of April 27, 2026.*