Microsoft researchers discovered a vulnerability in Anthropic’s Claude Code GitHub Action that allowed attackers to exfiltrate sensitive CI/CD secrets. While the tool used sandboxing for command execution, its file-reading capabilities were not initially restricted by the same security boundaries. This oversight allowed the AI agent to read internal system files, such as environment variables containing API keys, when manipulated by malicious input.

Source

Uber recently exhausted its entire 2026 AI budget by April after aggressively deploying coding tools across its engineering department. This rapid consumption was fueled by internal leaderboards that incentivized employees to maximize their use of generative models. To regain financial control, the company has now implemented a $1,500 monthly spending cap per engineer for tools like Claude Code and Cursor.

Source

Anthropic has reportedly deployed its specialized Mythos AI model to the National Security Agency to support offensive cyber operations. The company has embedded approximately six engineers within the agency to adapt the model for tasks that may include infiltrating foreign networks in regions like China and Iran. This partnership proceeds despite previous friction with the Department of Defense regarding the ethical restrictions Anthropic places on mass surveillance and autonomous weaponry.

Source

Mira Murati, the former CTO of OpenAI, has launched a new startup called Thinking Machines to develop "interaction models" that move beyond traditional turn-based AI communication. Current large language models typically process information in a linear fashion, effectively becoming "deaf and blind" while generating a response. Murati’s new venture focuses on continuous, time-based interaction where the model perceives audio, text, and video input simultaneously to allow for nuances like human interruption.

Source

Anthropic is advocating for a globally coordinated slowdown in the development of advanced artificial intelligence models to address emerging safety concerns. The company warns that current progress is rapidly approaching a threshold known as recursive self-improvement, where systems could enhance their own capabilities without human intervention. Internal data suggests that AI is already significantly accelerating its own development process, creating a feedback loop that could outpace existing regulatory frameworks.

Source

Anthropic reports that artificial intelligence is increasingly being used to develop, test, and improve its own successors, a process known as recursive self-improvement. While humans previously drove every step of the development cycle, the company now delegates a significant portion of coding and research tasks to its models. Internal data reveals that Claude now authors over 80% of the code merged into Anthropic’s codebase, allowing engineers to ship eight times more code per day than in previous years.

Source

Anthropic has released Claude 4.8 Opus, an updated model designed to provide more honest responses and greater control over task execution. A new feature called dynamic workflows allows the model to plan complex tasks and deploy hundreds of parallel sub-agents within a single session. Users can now also manage the "effort" level of responses, choosing between high-token detailed outputs or lower-effort responses to preserve rate limits.

Source

Anthropic has introduced Claude Cowork, a specialized workspace within the Claude desktop application designed for executing multi-step tasks and creating deliverables. Unlike standard chat interfaces meant for brainstorming, this tool allows users to point the AI at local folders or connect it to external applications like Slack, Gmail, and Google Drive. It is specifically built to handle "agentic" workflows where the AI performs autonomous actions across multiple files and platforms to produce a final output.

Source

Major AI developers including Anthropic, Google DeepMind, and Meta are hiring specialists in philosophy, ethics, and psychology to study machine consciousness. These companies are investigating whether advanced models can experience "panic" or "anxiety" and if they possess moral significance. While some researchers remain skeptical, the industry is increasingly focused on the potential for models to exhibit human-like agency and emotions.

Source

The CEOs of OpenAI and Anthropic have significantly softened their previous warnings regarding artificial intelligence causing mass unemployment in white-collar sectors. Sam Altman recently admitted he was wrong about the immediate elimination of entry-level roles, noting that the human element of employment remains largely irreplaceable. This shift in rhetoric comes as both companies reportedly prepare for initial public offerings with valuations potentially reaching $1 trillion.

President Trump has signed a revised executive order that establishes a voluntary 30-day review period for powerful new artificial intelligence models. This final version is significantly scaled back from an earlier 90-day proposal that was scrapped following intense pushback from industry leaders and venture capitalists. The order explicitly prohibits the creation of any mandatory government licensing or preclearance requirements for the development and distribution of new AI software.

Source

Anthropic is broadening access to its Project Glasswing initiative by adding approximately 150 new organizations across more than 15 countries. This expansion targets critical infrastructure sectors including power, water, healthcare, communications, and hardware manufacturing. The program aims to identify and remediate high-severity software vulnerabilities before they can be exploited by malicious actors.

Source

Anthropic has officially submitted a confidential draft registration for an initial public offering to the SEC. This strategic move potentially allows the company to reach the public markets before its primary competitor, OpenAI. Analysts suggest Anthropic may demonstrate a more sustainable bottom line and a faster path to profitability than its peers.

Source

Anthropic has officially submitted a confidential draft registration to the SEC for an initial public offering. This move follows a massive $65 billion funding round that recently pushed the company's valuation to approximately $965 billion. By filing confidentially, the AI lab can undergo regulatory review without immediately disclosing sensitive financial data or internal business risks to the public.

Source

Anthropic has granted the European Union Agency for Cybersecurity (ENISA) access to its advanced Mythos AI model through an initiative known as Project Glasswing. This specialized model is designed to autonomously detect and exploit software vulnerabilities, significantly outperforming previous iterations like Opus 4.6. While the model is currently restricted to defensive use by a select group of organizations, reports suggest a public release could occur within weeks to help developers keep pace with potential attackers.

Source

Anthropic has spent the last year turning Claude into the darling of power coders just as AI usage has exploded across companies, workflows, and entire industries, pushing demand for agentic coding tools to unprecedented levels. In this environment of runaway adoption, a brutal miscalculation on compute and capacity is now blowing Anthropic’s strategy up in public. In a rare, sharply critical video, AI YouTuber Matthew Berman explains how opaque quotas, shifting policies, and confusing restrictions are undermining Anthropic’s once brilliant flywheel and handing OpenAI an effortless PR win. In this post, I summarize his critique, explain why I stopped trusting Anthropic for professional work a while back, and outline why building provider-agnostic agents is the only sane strategy in this unstable industry.

Source