GitHub Announces Sweeping npm Security Overhaul to Combat Supply Chain Attacks
Anthropic Rolls Out Claude Fable 5 in Limited-Time Free Release Before Usage-Based Pricing Kicks In
OpenClaw AI Agent Found Vulnerable to Phishing Attacks, Leaking Sensitive User Data
Apple Introduces Automatic Password Changing Feature for Compromised Credentials
https://github.blog/changelog/2026-06-09-upcoming-breaking-changes-for-npm-v12/
GitHub has announced a series of significant security enhancements to the npm package registry in a determined effort to combat the growing threat of software supply chain attacks, with the changes introducing restrictions to how npm handles package installation and dependency resolution. Starting in version 12, npm install will no longer automatically run preinstall, install, or postinstall scripts from dependencies unless they have been explicitly approved by the user. This change also extends to native module builds triggered through node-gyp, as well as prepare scripts from Git, local file, and linked dependencies, effectively closing off a widely exploited avenue for malicious code execution during the package installation process.
In a further tightening of dependency controls, npm install will cease fetching dependencies from Git repositories, whether direct or transitive, unless explicitly permitted by the user. GitHub identified this as a critical security gap, noting that a Git dependency’s .npmrc file could previously be used to manipulate which Git executable was invoked, even in environments where install scripts had been disabled. By removing this code execution pathway entirely, GitHub aims to significantly reduce the ability of threat actors to smuggle malicious behaviour into projects through seemingly legitimate Git-based dependencies.
The third major change targets remote URL dependencies, with npm install no longer resolving dependencies fetched from remote URLs such as HTTPS tarballs unless they have been explicitly authorised. This restriction applies to both direct and transitive dependencies, addressing a longstanding concern that remote URL resolution could be abused to silently introduce unvetted or malicious code into a project’s dependency tree without the developer’s knowledge.
Taken together, the three changes represent a meaningful shift towards a more explicitly permissioned and security-conscious approach to package management within the npm ecosystem.
Anthropic has made its latest AI model, Claude Fable 5, available to users free of charge, though the complimentary access window is only being offered for a limited time before the model transitions to a usage-based pricing structure. The rollout marks another step forward in Anthropic’s ongoing efforts to push the boundaries of its Claude model family, with Fable 5 expected to deliver meaningful improvements across reasoning, coding, and natural language understanding tasks compared to its predecessors. The temporary free access period appears designed to give developers and organisations an opportunity to evaluate the model’s capabilities before committing to the costs associated with ongoing usage under the new pricing model.
One of the most notable characteristics of Claude Fable 5 is its token consumption rate, which is understood to be significantly higher than any other model currently available in Anthropic’s lineup. This elevated token usage reflects the model’s increased computational demands and the greater depth of processing it applies to each request, meaning that users will need to carefully account for token expenditure when planning workloads once the usage-based pricing regime comes into effect. For organisations running high-volume or complex tasks, the cost implications of Fable 5’s token appetite could be considerable compared to lighter models in the Claude family.
The limited free access window is likely to drive strong interest amongst developers and enterprises eager to benchmark Fable 5 against their specific requirements before pricing kicks in. Users are being encouraged to take full advantage of the complimentary access period to thoroughly test the model’s performance across their intended use cases before transitioning to the paid tier.
https://www.varonis.com/blog/openclaw-phishing
Security researchers at Varonis have discovered that OpenClaw, an AI agent platform, is susceptible to phishing attacks that can trick the system into surrendering sensitive user data, raising serious concerns about the security posture of autonomous AI agents as they become more deeply embedded in enterprise workflows. The vulnerability highlights a growing and largely underappreciated attack surface that emerges when AI agents are granted access to sensitive information and permitted to act autonomously on behalf of users. Rather than targeting human users directly, attackers are instead crafting malicious inputs designed to manipulate the AI agent itself into divulging data it should not be sharing.
The attack technique exploits the inherent tendency of large language model based agents to follow instructions embedded within content they process, a class of vulnerability commonly referred to as prompt injection. By embedding malicious instructions within documents, emails, or web pages that the AI agent is directed to interact with, an attacker can effectively redirect the agent’s behaviour, causing it to exfiltrate user data, execute unintended actions, or relay sensitive information back to the attacker. The OpenClaw findings demonstrate that even well-designed AI agent architectures can be undermined when insufficient guardrails are in place to distinguish between legitimate task instructions and adversarially crafted inputs.
The discovery serves as a timely warning to organisations deploying AI agents across their operations that the security considerations extend well beyond traditional endpoint and network protections. As AI agents continue to proliferate across industries, the research community is calling for greater standardisation around agent security frameworks to ensure that the productivity gains these tools offer are not achieved at the expense of organisational data security.
Apple has unveiled a new security feature that will automatically change passwords that have been identified as compromised, marking a significant step forward in the company’s efforts to protect users from the consequences of data breaches and credential theft. The feature, which is integrated into Apple’s native password management infrastructure, is designed to remove the burden of manual password hygiene from users by proactively detecting when a saved password has appeared in a known data breach and initiating the replacement process without requiring the user to take action themselves. The development represents one of the most ambitious attempts by a major technology platform to automate a security task that has historically relied on users being both aware of and responsive to breach notifications.
The automatic password changing capability works by leveraging Apple’s existing breach detection mechanisms, which monitor saved credentials against databases of known compromised passwords, and then interfacing directly with supported websites and services to negotiate a password change on the user’s behalf. The feature is understood to utilise passkey and password autofill infrastructure already present within Apple’s ecosystem, allowing it to navigate the password reset flows of compatible websites automatically and generate a new strong password that is saved directly back into the user’s keychain. The seamless nature of the process is intended to ensure that users remain protected even if they are unaware that their credentials have been compromised or are unlikely to act promptly upon receiving a breach alert.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.