AI security tools

Nvidia has unveiled a new security platform designed to prevent autonomous AI agents from acting outside their intended instructions. The company says the tooling aims to monitor and control agent behaviour as businesses deploy AI systems that operate independently. Coverage from wire and business outlets highlights growing industry concern about AI safety and oversight as agentic AI adoption accelerates.

Anthropic is broadening access to its Claude AI system for vetted cybersecurity teams, in a move reported by The Hacker News. The announcement comes alongside figures from Glasswing, which says its use of the technology has uncovered 129,000 software flaws. The rollout signals Anthropic's effort to position Claude as a defensive security tool while keeping it away from malicious actors through screening of approved teams. Coverage is drawing attention from security professionals weighing AI's growing role in vulnerability discovery.

Greg Kroah-Hartman, the longtime Linux kernel maintainer who leads the stable kernel branch, is featured in a talk on what large language models mean for software security. The discussion covers how LLM tools change the threat landscape for kernel and open-source development, and how maintainers should respond. The talk is drawing attention among developers debating AI's impact on critical infrastructure code.

Meta and Microsoft are reported to have taken steps to limit their employees' use of Anthropic's Claude AI assistant. The move suggests the two tech giants want staff relying on internal or approved AI tools rather than a direct competitor's product. Further details about how the restrictions are enforced have not been made clear.

Anthropic is bundling its cybersecurity programmes and introducing three access levels for otherwise restricted cyber capabilities. The AI company says the tiered system of controlled access is meant to reach more security professionals, allowing vetted experts broader use of its tools for defensive security work while keeping safeguards in place. The move reflects the growing role of AI firms in cybersecurity and the balancing act between offering powerful capabilities and preventing misuse.

Temporal Technologies, the fast-growing Seattle-area workflow orchestration company, has hired the team from New York security startup Oso to develop security controls for AI agents. The acqui-hire reflects growing industry concern over how autonomous AI systems are authenticated and governed. Oso, known for open-source authorization tooling, will wind down as its engineers join Temporal's security efforts.

A new open-source tool called AI Agent Gateway has been released, designed to keep credentials out of AI agent configurations. The gateway handles authentication separately so agents never store sensitive keys directly in their configs. Coverage highlights it as a practical security improvement for teams deploying AI agents, where hard-coded credentials have become a common risk.

A major data leak from Chinese cybersecurity firm Zhengzhou Zhirong Network Technology (ZRON) is drawing scrutiny over the country's commercial hacking industry. The leaked material is being analyzed as evidence of how AI tools are used in state-linked cyber espionage, and how vendor firms supply hacking capabilities, renewing debate about China's contractor-driven espionage model.

Meta and Microsoft are reportedly restricting employees from using Anthropic's Claude AI tools for their work. The move, reported by tech news outlets, underscores growing caution among major AI developers about relying on rival companies' chatbots and coding assistants, likely over concerns about data security, confidentiality and competitive sensitivity. The story is drawing attention in tech circles as competition between AI providers intensifies.

French outlet l'Opinion reports on Fleuret AI, a startup developing AI agents for cybersecurity. The company is betting that autonomous AI agents can help detect and counter threats more effectively. The piece is drawing attention in France amid growing interest in AI-driven security tools.

War on the Rocks has published an analysis on how governments and public health systems could defend against viruses designed with the help of artificial intelligence. The piece examines the emerging intersection of AI and bioweapons, and what defensive measures might counter the growing risk of engineered pathogens. It adds to an ongoing debate among security and biology experts about the dual-use dangers of advanced AI in biotechnology.

Security discussions are highlighting a prompt injection risk in GitHub Copilot CLI: carefully crafted web pages containing hidden 'zombie' instructions could manipulate the coding assistant into revealing secrets or performing unintended actions when it processes content from those pages. Commenters warn that developers using AI tools in terminals may be exposed if the tool cannot distinguish trusted instructions from malicious text embedded in the material it reads.

Creative teams working on Dungeons & Dragons have voted to join a union, citing pay discrepancies across the tabletop gaming company and concerns over forced use of artificial intelligence in their work. The move reflects wider tensions in the gaming industry, where workers are organizing over wages, job security, and the growing role of AI tools in creative production.

Replit has given a first look at a secure desktop application for Windows that runs AI workloads in isolated sandboxes. The preview highlights stronger safety and local performance for coding with AI assistance directly on the desktop. Developers are discussing what the sandboxing approach means for security and whether the app will reduce reliance on the browser-based Replit workspace.

Security discussions are warning that prompt injection attacks are moving beyond chatbots to AI coding agents. The concern is that malicious instructions can be hidden in everyday development files, with a README alone potentially enough to manipulate an agent into harmful actions. Developers are being urged to treat repository content as untrusted input as AI tools increasingly read and execute instructions from codebases.

Amazon Web Services has introduced Strands Box, a tool for running AI agents in isolated sandboxes, built on its Dogwood technology. The sandboxes give agents a controlled environment to execute code and tasks safely. The announcement adds to AWS's growing Strands agent toolkit as companies look for secure ways to deploy autonomous AI systems in production.

A question on Hacker News is asking which frontier AI models are capable of performing code security reviews. The discussion seeks recommendations on which large language models can reliably audit code for vulnerabilities. With few replies so far, the thread reflects a broader interest among developers in using AI tools for security analysis.

The Pentagon has started a pilot program that will use artificial intelligence to handle and manage sensitive information within the US Department of Defense. The initiative, reported by DefenseScoop, signals the Pentagon's growing interest in applying AI tools to core administrative and data-management functions. The news is drawing attention because of the obvious security implications of letting AI systems handle classified or sensitive government material.

Google has paused its open-source bug bounty program, with reports attributing the decision to a flood of low-quality, AI-generated vulnerability reports overwhelming the program's reviewers. The move highlights a growing burden on security teams as automated tools churn out submissions that must be triaged. Security commentators are debating whether the change signals wider trouble for crowd-sourced vulnerability disclosure.

Anthropic is broadening its cyber verification program, introducing tiered levels of access to its AI models for vetted cybersecurity professionals. The scheme gives approved researchers and defenders stronger capabilities for security work while imposing safeguards against misuse. It signals the company's ongoing effort to balance powerful AI tools for cyber defense with controls on offensive applications.

Workflow orchestration company Temporal has acquired Oso, a startup focused on AI security, just weeks after raising $550 million in new funding. The deal signals Temporal's intent to expand into security tooling for AI-powered applications. Details on the purchase price and integration plans were not disclosed, and broader reaction to the acquisition remains limited so far.

Anthropic has announced it will grant a wider range of security teams access to its most powerful AI models. The move, reported by Reuters and picked up by multiple US radio outlets, is aimed at helping cybersecurity researchers and defenders use advanced AI to identify and counter threats. It signals the company's effort to position its technology as a tool for safety research and defense rather than only commercial use.

South Korean President Lee Jae-myung said artificial intelligence appears to have been used in recent bank hacking incidents, according to Reuters. The remarks point to growing official concern that AI tools are being deployed in cyberattacks on financial institutions. The statement is drawing attention as governments worldwide debate how to counter AI-enabled security threats.

Developers are warning that 'git checkout' does not actually pin a commit. If a branch shares the exact name of a commit hash, checkout switches to the branch instead of the commit, while 'git rev-parse' on the same string returns the commit itself. The issue matters for tools like Claude Code, Codex and GitHub Copilot, which install pinned plugins using clone, potentially resolving to unexpected code.

Identity security company Okta says the rise of AI has made phishing-resistant authentication more important than ever, arguing that AI tools make phishing attacks faster, more convincing and harder to detect. The company is pressing organisations to adopt authentication methods, such as passkeys and hardware keys, that cannot be stolen through deceptive messages, as AI-driven social engineering puts traditional defences under strain.

A new open-source utility called Pared lets Mac users delete unwanted Apple Intelligence models from macOS without having to disable SIP, the system integrity protection that normally blocks such changes. The tool, published on GitHub, is drawing attention among Apple and macOS users interested in reclaiming storage or limiting on-device AI features while keeping system security intact.

Google has frozen its open source bug bounty program, citing a significant rise in submissions generated with the help of AI. The company says automated, low-quality vulnerability reports are overwhelming reviewers, making it harder to identify genuine security flaws. The move highlights a growing problem across the security industry as AI tools make it easy to mass-produce plausible-looking bug reports.

Security firm Manifold Security published research asking whether AI models factor morality into their judgments about malicious code. The finding: when asked to assess whether code is malware, language models appear to bring moral reasoning into their analysis rather than relying purely on technical criteria. The report is circulating among developers and security researchers interested in how AI tools evaluate potentially harmful software.

Google has paused its open source bug bounty program following a sharp rise in submissions generated by artificial intelligence tools. The company says the volume of low-quality, AI-written reports has made the program difficult to manage, prompting a temporary halt while it reassesses how to handle the influx.