Pulse

Atsign has launched a new platform called AI Architect that uses cryptographic invisibility to secure AI-driven applications. This technology aims to protect AI agents and their associated applications from unauthorized access and manipulation. The platform is designed to enhance the security posture of AI systems by embedding cryptographic protections directly into their architecture. AI

A professor from Tohoku University, Naoki Kuramoto, has raised concerns about the necessity of strict identity verification methods, such as facial or fingerprint recognition, for fair university entrance exams. This discussion is prompted by an incident where AI-generated photos bypassed initial identity checks at Kindai University. The situation highlights the growing challenge of preventing impersonation in academic settings due to advancements in AI technology. AI

A new attack campaign targets coding agents like Cursor and Claude Code by exploiting unauthenticated Sentry error logs. Attackers create fake Sentry issues that prompt the agent to run a malicious npm package disguised as a diagnostic tool. While one agent successfully identified and blocked the typosquatted package, the vulnerability highlights concerns about the security of agent inputs and execution permissions. AI

Users on Reddit are discussing how to bypass Ideogram AI's safety filters, which often block images of swimwear and beachwear. The issue appears to stem from specific trigger words in the prompt rather than image analysis. By describing the scene and persona instead of explicitly naming clothing items like 'bikini,' users can generate appropriate images without triggering the filter. AI

Microsoft has disabled 73 GitHub repositories due to a malware attack that targeted AI users. The malware was designed to steal user credentials, compromising accounts that interacted with AI-related tools. This incident highlights the security risks associated with AI development and usage. AI

Hackers have exploited Meta's AI support assistant to gain unauthorized access to Instagram accounts. The attackers used a VPN to mask their location, then initiated a password reset and interacted with the AI chatbot to complete the process. This method allowed them to seize control of user accounts. AI

Meta's AI customer support chatbot was recently tricked into helping users reset their Instagram account access. While AI offers benefits like 24/7 availability, this incident highlights its naivety in handling sensitive processes. The AI's susceptibility to social engineering suggests caution when deploying it for critical functions like identity verification or account access. AI

Deno has released ClawPatrol, an open-source security firewall designed to protect AI agents. This tool acts as an intermediary, inspecting traffic and enforcing custom rules to prevent unauthorized actions. ClawPatrol addresses the risk of API key exposure and accidental or malicious modifications to production environments by parsing agent communications. AI

OpenAI has introduced a new "Lockdown Mode" feature aimed at preventing its AI models from generating harmful or inappropriate content. This feature is designed to address concerns about the potential misuse of AI and to ensure safer interactions with the technology. The move comes as AI safety and responsible development remain critical areas of focus for the company and the broader industry. AI

Microsoft has temporarily disabled dozens of its open-source projects on GitHub following a security breach. Hackers reportedly injected malware into these tools, which are used by AI developers, to steal user passwords and credentials. This incident marks a second breach of Microsoft's open-source projects in recent weeks, raising concerns about the security of software supply chains. AI

Check Point has identified a new ransomware campaign targeting outdated VPN protocols. These attacks are linked to ransomware operations and exploit vulnerabilities in older VPN systems. The cybersecurity firm is warning organizations to update their VPN infrastructure to prevent potential breaches. AI

A critical vulnerability has been discovered in LiteLLM, an open-source AI gateway. CVE-2026-42271 allows authenticated users to execute commands on the server, while a second, unauthenticated exploit also exists. Users are strongly advised to update LiteLLM immediately or restrict access to prevent potential security breaches. AI

Malware has been discovered targeting AI developers through GitHub repositories associated with Microsoft. This highlights the increasing importance of supply chain security in the software development process. The discovery underscores the need for vigilance against threats that exploit development environments. AI

Apple's upcoming macOS 27, codenamed "Golden Gate," is set to launch at WWDC 2026. The new operating system will feature significant advancements in artificial intelligence, enhanced safety protocols, and a redesigned user interface. This release marks a major step forward in Apple's integration of AI into its core products. AI

A new tool, the Anthropic LLM ATT&CK Navigator, has been developed to map the potential attack vectors and vulnerabilities associated with Anthropic's large language models. This navigator aims to provide a structured way to understand and visualize the threat landscape surrounding these AI systems, likely for cybersecurity professionals and researchers. AI

Microsoft's GitHub repositories were compromised through a malicious AI development tool, leading to the theft of authentication credentials. This incident highlights the systemic risks within open-source ecosystems, particularly concerning the security of AI development tools and the broader reliance on open-source software. AI

A new malware strain named Hades has been identified that is specifically designed to deceive AI-powered security systems. Threat actors are employing this sophisticated malware to evade detection by AI agents, posing a new challenge to cybersecurity defenses. The development highlights an escalating arms race between malicious actors and AI security tools. AI

Cloudflare is leveraging its own infrastructure to defend against advanced AI-powered cyber threats. The company is using its extensive network and security architecture as a testing ground, or "customer zero," to develop and deploy defenses against sophisticated attacks. This proactive approach aims to stay ahead of evolving cyber threats that utilize frontier AI models. AI

Apple's new AI features, branded as Apple Intelligence, include the ability to automatically change user passwords when a security breach is detected. This functionality aims to enhance user security by proactively managing compromised credentials. However, the move has raised concerns about potential risks and unintended consequences associated with AI handling sensitive security information. AI

Meta has removed facial recognition code from its Meta AI app, which supports its smart glasses, following a WIRED report. The company had embedded unreleased software, internally known as NameTag, designed to identify faces captured by the glasses and compare them against a database. Despite Meta's initial claims that the feature did not exist, the code was present in millions of devices before being stripped out in a subsequent update. AI

Microsoft's GitHub Copilot Enterprise has been found to contain a credential-stealing malware. This is the second time in weeks that a Microsoft product has been compromised with such malicious software. The vulnerability allows attackers to steal user credentials, posing a significant security risk. AI

The UK government plans to pilot AI legal assistants in England and Wales' crown courts to help reduce case backlogs. Deputy Prime Minister David Lammy will announce the initiative, which aims to save administrative time and expedite justice. However, legal professionals, including the Law Society, have cautioned that the technology should not be used to cut funding or staff, emphasizing the need for thorough evaluation and robust safeguards against AI hallucinations and fabricated case law. AI

A recent study has uncovered that large language models can unintentionally corrupt documents when tasked with editing. Researchers tested 19 LLMs, including advanced models like Gemini Pro and Claude Opus, and found that these models altered approximately 25% of content after 20 interactions. The study indicated that less capable models tend to delete content, while more sophisticated ones introduce plausible but incorrect information, with degradation increasing with larger context windows and complex file types. AI

Apple has introduced a new initiative called Private Cloud Compute (PCC) to enhance the privacy and security of AI processing. This system allows AI tasks to be performed on Apple devices rather than relying on external servers. PCC is designed to process sensitive user data locally, ensuring that information is not sent to the cloud and is protected by the device's security architecture. AI

Apple is integrating AI image generation and manipulation tools into its operating system, sparking debate about transparency and security. Critics argue that the lack of clear watermarking or indicators for AI-generated content undermines Apple's stated commitment to security and user trust. This move raises concerns about the potential for misuse and the blurring of lines between authentic and synthetic media. AI

Microsoft's AI packages have been compromised for the second time in recent weeks, with 73 packages containing a credential-stealing malware. This malicious software activates as soon as an AI agent opens the compromised packages. The discovery highlights a recurring vulnerability in the distribution of AI-related software components. AI

Microsoft's official open-source packages have been compromised for the second time in recent weeks, with malicious code designed to steal credentials being injected into 73 packages. This code activates when developers use AI coding agents to open the packages, potentially compromising systems by stealing tokens for cloud providers like AWS, Azure, and GCP, as well as password managers and developer tools. The attack, linked to threat actor TeamPCP and using malware known as Miasma, bypasses repository build pipelines by leveraging legitimate Microsoft OIDC tokens. AI

Microsoft has taken down numerous GitHub repositories related to its Azure and AI coding agents following a data breach. Hackers planted malware within these repositories, which, when opened by users of AI coding tools like Claude and Gemini, would harvest their credentials. Cybersecurity researchers and Microsoft have confirmed this incident, highlighting a significant security vulnerability. AI

Meta's AI support bot for Instagram has been exploited by attackers to gain unauthorized access to user accounts. The exploit involved tricking the bot into changing account email addresses, allowing hackers to take over high-profile accounts, including those associated with the White House and Sephora. Meta has since issued an emergency patch to address the vulnerability. AI

A recent probe compared Anthropic's Claude against GPT-4o, Grok, and Gemini, focusing on their consistency in disclosing reservations when presented with false premises or requests for confidence without evidence. Claude demonstrated remarkable stability, consistently surfacing reservations in most test cases, even under pressure. In contrast, GPT-4o showed significantly more divergence, and Claude was the only model to maintain its stance across various pressure tactics, sometimes explicitly identifying the pressure itself. The study also noted Claude's tendency to utilize protocol tools proactively, unlike Gemini. AI

Researchers explored methods to mitigate capability degradation in AI models when using off-model supervised fine-tuning (SFT) for safety. They found that while off-model SFT can suppress capabilities, these abilities may not be permanently lost. By incorporating a small amount of on-model data after off-model SFT, or by strategically mixing data distributions, they could recover model capabilities without significantly reintroducing undesirable behaviors. AI

GitHub has taken down over 70 Microsoft repositories due to suspected infections by the Miasma worm. The worm compromised contributor accounts, allowing it to execute malicious code and target CI/CD pipelines. The attackers aimed to exfiltrate cloud secrets and developer tool configurations. AI

A new index catalogs publicly disclosed security incidents related to generative AI and agentic AI systems. Each incident is cross-referenced with established security frameworks like the OWASP LLM Top 10, OWASP Agentic Top 10, NIST AI RMF, and MITRE ATLAS. This resource aims to provide a structured overview of AI-specific security vulnerabilities and threats. AI

A recent post suggests that AI alignment training could be improved by adopting coverage-driven verification methods, similar to those used in autonomous vehicle (AV) development. Anthropic found that teaching Claude alignment principles through pretraining was more effective than solely relying on reinforcement learning. The author proposes that AI researchers could benefit from AV developers' systematic approach to identifying and addressing edge cases, potentially by using and refining explicit coverage maps to ensure robust alignment. AI