Claude 4 Opus
Claude 4 Opus is Anthropic’s most capable model for long-horizon agentic work, multi-step reasoning, and complex coding projects.
Released
May 22, 2025
Type
reasoning
License
proprietary
Context
200,000 tokens
Pricing
Input
$15.00 / 1M tokens
Output
$75.00 / 1M tokens
Benchmarks
| SWE-bench Verified | 72.5% |
| GPQA Diamond | 79.6% |
Capabilities
Links
Claude 4 Opus in the news
CoinTelegraph · Jul 29, 2026
MoonPay vault enables ChatGPT and Claude users to authorize crypto transactions
MoonPay has launched its PayBox vault, a new feature designed to enable users of large language models like OpenAI's ChatGPT and Anthropic's Claude to authorize cryptocurrency transactions. This integration allows users to conduct purchases, token swaps, and cross-chain transfers directly through the interfaces of these AI assistants, while crucially retaining full custody of their digital assets. The PayBox vault acts as an intermediary, facilitating these operations without requiring users to directly expose their private keys or connect their wallets to third-party applications. This development signifies a growing trend of integrating decentralized finance (DeFi) functionalities into mainstream digital platforms. By embedding crypto transaction capabilities within AI chatbots, MoonPay aims to lower the barrier to entry for cryptocurrency adoption, making it more accessible and intuitive for a broader audience. Users can leverage the conversational interfaces of ChatGPT and Claude to initiate and approve transactions, a process that traditionally involves navigating complex wallet interfaces and understanding technical jargon. The emphasis on self-custody means that users maintain control over their private keys, mitigating risks associated with centralized exchanges or custodial services. MoonPay, a company specializing in fiat-to-crypto on-ramps and off-ramps, has been actively developing solutions to streamline the user experience in the cryptocurrency space. The PayBox vault represents a significant step in this direction, by bridging the gap between natural language interaction and blockchain operations. The ability to authorize transactions via AI assistants could lead to more seamless and context-aware financial interactions. For instance, a user might ask ChatGPT to "buy 0.1 Bitcoin" or "swap 5 ETH for USDC," and the PayBox vault would then facilitate the secure authorization of this request, provided the user has sufficient funds and has pre-authorized MoonPay's service. The integration is expected to benefit both novice and experienced cryptocurrency users. For beginners, it offers a more guided and less intimidating entry point into managing digital assets. For advanced users, it provides an additional layer of convenience and a novel way to interact with their portfolios. The underlying technology of the PayBox vault is designed to ensure that sensitive information, such as private keys, never leaves the user's control. Instead, it facilitates the signing of transactions on the blockchain through secure protocols, ensuring that the user's assets remain protected. This approach aligns with the core principles of decentralization and user empowerment in the crypto ecosystem.
The Hacker News · Jul 29, 2026
Ruflo MCP Flaw Lets Unauthenticated Attackers Run Commands and Poison AI Memory
A critical security vulnerability, designated CVE-2026-59726 with a perfect CVSS score of 10.0, has been identified in Ruflo, an open-source agent meta-harness. This flaw permits unauthenticated attackers to achieve remote code execution and potentially poison the memory of artificial intelligence models. The vulnerability affects all versions of Ruflo prior to version 3.16.3. Noma Security's researchers have codenamed this critical issue RufRoot. Ruflo serves as a meta-harness, meaning it is a framework designed to manage and orchestrate multiple AI agents. Specifically, it is utilized with AI models such as Anthropic's Claude Code and OpenAI's Codex, which are specialized for code generation and understanding. The meta-harness allows developers to build complex AI applications by chaining together different AI agents, enabling them to perform more sophisticated tasks. The RufRoot vulnerability exploits weaknesses in how Ruflo handles user inputs and agent interactions, allowing an attacker to inject malicious commands that are then executed by the underlying AI agents or the system hosting them. This could lead to a complete compromise of the affected system. The ability to poison AI memory means that an attacker could subtly alter the AI's learned knowledge or operational state, leading to incorrect outputs, biased decision-making, or further exploitation. This is particularly concerning for AI systems that rely on accurate and untainted memory for their functionality, such as those used in code development, data analysis, or autonomous operations. The severity of the CVSS score of 10.0 indicates that the vulnerability is exploitable remotely, requires no authentication, and has a high impact on confidentiality, integrity, and availability. The researchers at Noma Security disclosed the vulnerability and its implications, emphasizing the urgent need for users to update their Ruflo installations. The project's maintainers have released version 3.16.3 to address this critical flaw. Users of Ruflo are strongly advised to upgrade immediately to mitigate the risk of exploitation. The disclosure of RufRoot highlights the ongoing security challenges in the rapidly evolving field of AI agent development, where complex integrations and open-source components can introduce unforeseen vulnerabilities. The reliance on meta-harnesses like Ruflo for orchestrating powerful AI models necessitates rigorous security auditing and prompt patching of discovered flaws to maintain the integrity and security of AI systems and the data they process.
Decrypt · Jul 29, 2026
Morning Minute: Claude Mythos Breaks Post-Quantum Cryptography
Anthropic's Claude Mythos AI model has successfully demonstrated the ability to break post-quantum cryptography, a significant development that could have profound implications for digital security. The breakthrough was detailed in a recent announcement, highlighting the model's advanced capabilities in deciphering encryption methods designed to withstand attacks from future quantum computers. Post-quantum cryptography (PQC) refers to cryptographic algorithms that are thought to be secure against attacks by both classical and quantum computers. The development of these algorithms has been a major focus for cryptographers and cybersecurity experts worldwide, as the advent of powerful quantum computers could render current encryption standards obsolete. Claude Mythos, developed by Anthropic, is a large language model known for its sophisticated reasoning and problem-solving abilities. Its success in breaking PQC suggests that the threat posed by quantum computing to current security infrastructure may be more immediate than previously anticipated. This capability raises urgent questions about the readiness of global digital systems to transition to quantum-resistant encryption. The implications extend to sensitive data, financial transactions, government communications, and national security, all of which rely heavily on the integrity of current cryptographic protocols. While the specifics of how Claude Mythos achieved this breakthrough have not been fully disclosed, the achievement underscores the rapid advancements in artificial intelligence and its potential to disrupt established technological paradigms. Cybersecurity agencies and governments are expected to reassess their timelines and strategies for implementing quantum-resistant solutions. The race to develop and deploy PQC has been ongoing, with organizations like the U.S. National Institute of Standards and Technology (NIST) leading standardization efforts. However, this development may necessitate an acceleration of these efforts and a re-evaluation of the security assumptions underpinning much of the digital world. Beyond the headline news of Claude Mythos, other market and technological developments are also noteworthy. Markets are showing mixed performance ahead of the Federal Open Market Committee (FOMC) meeting, indicating investor caution. Additionally, new Wall Street giants are reportedly backing the Clarity Act, a piece of legislation likely aimed at enhancing financial market transparency or regulation. In the cryptocurrency space, Zcash has officially launched its Ironwood upgrade, a significant event for the privacy-focused digital currency, which is expected to introduce new features and improvements to its network's security and functionality.
Fast Company · Jul 29, 2026
Should AI companies be able to outsource safety?
The debate surrounding artificial intelligence safety and regulation has intensified following recent actions by the U.S. government and the release of advanced AI models by foreign companies. In June, the U.S. government mandated that Anthropic restrict access to its two newest AI models for foreign nationals, citing national security and cybersecurity concerns. This directive led Anthropic to temporarily revoke access for all users due to the inability to verify user nationalities in real-time. Shortly thereafter, the Chinese company Moonshot AI launched its Kimi K3 model and subsequently published its complete model weights. These events have brought to the forefront a long-standing discussion: the extent to which governments can impose AI safety requirements without stifling innovation or allowing international competitors to gain an advantage. The current discourse, however, often overlooks the intricate nature of AI product development, which involves a network of interconnected firms rather than a single, monolithic entity. My academic research indicates that regulatory frameworks can significantly reshape this development process by dictating which parties are responsible for investing in safety measures and which can defer such investments to others. These considerations are particularly pertinent as the European Commission prepares to implement key provisions of the EU AI Act. This legislation establishes distinct safety and transparency obligations for providers of general-purpose AI models, such as OpenAI and Anthropic, and for companies that develop AI applications, including voice assistants and customer service chatbots. As EU regulators begin enforcing these rules, they must carefully consider how requirements targeting one segment of the AI development chain will influence other parts of the ecosystem. For instance, a company developing software to summarize physicians' clinical notes would likely begin with a general-purpose AI model from a major developer like OpenAI or Anthropic. This foundational model would then be adapted by the medical software company for its specific clinical application. Within this layered development process, investments in safety are divided. The provider of the general-purpose model faces decisions related to model training, general evaluation, the implementation of safeguards, and comprehensive documentation. The medical software company, in turn, confronts a different set of safety challenges and responsibilities related to the application of the AI in a sensitive domain like healthcare. The EU AI Act's tiered approach, distinguishing between model providers and application developers, highlights the need for a nuanced understanding of how safety responsibilities are distributed across the AI value chain. The challenge lies in designing regulations that effectively promote safety without inadvertently creating loopholes or imposing undue burdens that could hinder the progress of AI development, especially in a globally competitive landscape.
Decrypt · Jul 28, 2026
Claude Mythos Cracked Post-Quantum Cryptography That Humans Spent Years Failing to Break
Anthropic's advanced AI model, Claude, has successfully identified a significant vulnerability in a post-quantum signature scheme, a development that could have substantial implications for future digital security. This particular scheme is currently being considered for U.S. federal standardization, highlighting the immediate relevance of this breakthrough. The AI's discovery represents a novel attack vector that human cryptographers had reportedly spent years attempting to uncover without success. The specific signature scheme targeted by Claude is known as CRYSTALS-Dilithium, which is one of the algorithms selected by the U.S. National Institute of Standards and Technology (NIST) as part of its initiative to standardize post-quantum cryptography. NIST's goal is to develop cryptographic standards that are resistant to attacks from future quantum computers, which are predicted to be capable of breaking many of today's widely used encryption methods. CRYSTALS-Dilithium is designed to provide digital signatures, a crucial component for verifying the authenticity and integrity of digital communications and transactions. Anthropic's research, detailed in a recent publication, outlines how Claude was able to find an attack that exploits a weakness in the mathematical underpinnings of the CRYSTALS-Dilithium algorithm. The nature of this attack is not fully disclosed in the initial reports, but it is described as a method that could potentially allow an adversary to forge signatures or compromise the integrity of data protected by the scheme. The fact that an AI model achieved this feat, surpassing the efforts of human experts over an extended period, underscores the growing capabilities of artificial intelligence in complex scientific and security domains. This discovery by Claude raises critical questions about the security of cryptographic standards currently under development and the role of AI in both offensive and defensive cybersecurity. While the development of quantum-resistant cryptography is a necessary step to secure future digital infrastructure, the rapid progress of AI in finding vulnerabilities suggests a continuous arms race between AI-powered code-breaking and AI-assisted code-making. Anthropic's work demonstrates that AI can be a powerful tool for cryptanalysis, potentially accelerating the process of identifying and rectifying weaknesses in cryptographic systems before they can be exploited by malicious actors. The company has indicated that it is working with NIST and other relevant bodies to share its findings and contribute to the ongoing efforts to secure the digital world against future threats.