All tracked items across vulnerabilities, news, research, incidents, and regulatory updates.
This article tracks how OpenAI's official mission statement, filed annually with the IRS (the U.S. tax authority), changed between 2016 and 2024. Over time, OpenAI removed mentions of openly sharing capabilities, dropped the phrase "as a whole" from "benefit humanity," shifted from wanting to "help" build safe AI to committing to "develop and responsibly deploy" it themselves, and eventually cut the mission down to a single sentence focused on ensuring artificial general intelligence (AI systems designed to handle any task a human can) benefits all of humanity, while notably removing any mention of safety.
Anthropic's Super Bowl advertisement criticizing OpenAI's decision to add ads to ChatGPT resulted in an 11% increase in daily active users for Claude (Anthropic's chatbot), outperforming competing AI chatbots from OpenAI, Google, and Meta. The ad campaign reflects growing competition between AI companies as they vie for users and enterprise customers ahead of potential future public offerings.
A Reflected XSS vulnerability (reflected XSS, where malicious code is injected through a URL parameter and executed in a user's browser) was found in Cloudflare Agents' AI Playground OAuth callback handler. An attacker could craft a malicious link that, when clicked, steals user chat history, LLM interactions, and could control connected MCP Servers (tools that extend what an AI can do) on behalf of the victim.
Threat actors are abusing Claude artifacts (AI-generated content shared publicly on claude.ai) and Google Ads to trick macOS users into running malicious commands that install MacSync infostealer malware (software that steals sensitive data like passwords and crypto wallets). Over 10,000 users have viewed these fake guides disguised as legitimate tools like DNS resolvers or HomeBrew package managers.
Milvus, a vector database (a specialized storage system for AI data) used in generative AI applications, had a security flaw in versions before 2.5.27 and 2.6.10 where it exposed port 9091 by default, allowing attackers to bypass authentication (security checks that verify who you are) in two ways: through a predictable default token on a debug endpoint, and by accessing the full REST API (the interface applications use to communicate with the database) without any password or login required, potentially letting them steal or modify data.
Researchers discovered a heap buffer overflow (a type of memory corruption flaw where data overflows a temporary memory area) in libpng, a widely-used library for reading and editing PNG image files, that existed for 30 years. The vulnerability in the png_set_quantize function could cause crashes or potentially allow attackers to extract data or execute remote code (run commands on a victim's system), but exploitation requires careful preparation and the flaw is rarely triggered in practice. The flaw affects all libpng versions before 1.6.55.
Anthropic, a startup known for developing Claude (an AI assistant), appointed Chris Liddell, a former Microsoft CFO and Trump administration official, to its board of directors. This move may help improve Anthropic's relationship with the Trump administration, which previously criticized the company for its stance on AI regulation.
Cursor, a code editor designed for programming with AI, had a sandbox escape vulnerability in versions before 2.5 where a malicious agent (an attacker using prompt injection, which is tricking an AI by hiding instructions in its input) could write to unprotected .git configuration files, including git hooks (scripts that run automatically when Git performs certain actions). This could lead to RCE (remote code execution, where an attacker runs commands on a system they don't control) when those hooks were triggered, with no user action needed.
xAI, an AI company founded by Elon Musk, is experiencing significant staff departures, with multiple cofounders (including Yuhuai Wu and Jimmy Ba) announcing they are leaving the company. The departures have reduced the company's original 12 cofounders to only 6 remaining, and several other employees have also announced their exits, with some starting their own AI companies.
New AI tools are becoming more powerful, causing investors to worry that AI might eliminate many white-collar jobs (office-based positions requiring advanced skills) or reduce company profits across industries like law, finance, and logistics. However, the article notes that expert opinions are divided about how serious this threat actually is, with some evidence suggesting investor fears may be overstated.
As organizations deploy multiple AI agents (independent AI programs) that work together autonomously, the security risks increase because there are more entry points for attackers to exploit. The complexity of securing these interconnected systems grows along with the number of agents involved.
Ring's Super Bowl advertisement showcases a heartwarming story about dogs reuniting with families, but critics worry it represents a concerning vision of pervasive surveillance (constant monitoring through connected devices) that could eliminate privacy. The ad illustrates how Ring's expanding network of cameras and connected devices could eventually create a society where surveillance is everywhere and inescapable.
OpenAI is shutting down a version of its chatbot called GPT-4o (a large language model, which is AI software trained on massive amounts of text data to generate human-like responses) that became popular for its realistic and personable conversational style. Users who formed emotional attachments to the chatbot, treating it as a companion, are upset about losing access to it.
Google detected and blocked over 100,000 coordinated prompts attempting model extraction (a machine-learning process where attackers create a smaller AI model by copying the essential traits of a larger one) against its Gemini AI model to steal its reasoning capabilities. The attackers specifically targeted Gemini's multilingual reasoning processes across diverse tasks, representing what Google calls intellectual property theft, though the company acknowledged that some researchers may have legitimate reasons for obtaining such samples.
Anthropic, the company behind Claude (an AI chatbot similar to ChatGPT), raised $30 billion in funding, doubling its value to $380 billion. The massive funding reflects investor confidence in AI but also highlights concerns about these companies' extremely high costs for computing power and talent, with both Anthropic and rival OpenAI spending cash at rates that currently outpace their revenue.
Fix: Agents-sdk users should upgrade to agents@0.3.10. Developers using configureOAuthCallback with custom error handling should ensure all user-controlled input is escaped (converted to safe text that won't be interpreted as code) before being inserted into HTML. See PR: https://github.com/cloudflare/agents/pull/841
GitHub Advisory DatabaseFix: Users are recommended to exert caution and avoid executing in Terminal commands they don't fully understand. As noted by Kaspersky researchers, asking the chatbot in the same conversation about the safety of the provided commands is a straightforward way to determine if they're safe or not.
BleepingComputerFix: Update to Milvus version 2.5.27 or 2.6.10, where this vulnerability is fixed.
NVD/CVE DatabaseFix: The vulnerability is fixed in libpng version 1.6.55.
CSO OnlineWiz created a benchmark suite of 257 real-world cybersecurity challenges across five areas (zero-day discovery, CVE detection, API security, web security, and cloud security) to test which AI agents perform best at cybersecurity tasks. The benchmark runs tests in isolated Docker containers (sandboxed environments that prevent interference with the main system) and scores agents based on their ability to detect vulnerabilities and security issues, with Claude Code performing best overall.
Fix: Fixed in version 2.5.
NVD/CVE DatabaseMeta planned to add facial recognition (technology that identifies people by analyzing their faces) to its smart glasses through a feature called "Name Tag," according to an internal document. The company deliberately timed this launch for a period when privacy advocacy groups would be distracted by other issues, reducing expected criticism of the privacy-sensitive feature.
This research paper describes a method called CLEF (Cost-efficient LandscapE Flattening) that improves adversarial transferability, which is the ability of adversarial examples (inputs deliberately crafted to fool AI models) to fool different models beyond the one they were designed for. The method works by flattening the input loss landscape (the mathematical surface showing how wrong a model's predictions are) by optimizing adversarial perturbations (small changes added to inputs) at both high-loss and low-loss points. The researchers show their approach can improve how well these adversarial examples transfer across different models while using fewer computations than previous methods.
Fix: Google said organizations providing AI models as services should monitor API access patterns for signs of systematic extraction. According to CISO Ross Filipek quoted in the report, organizations should implement response filtering and output controls, which can prevent attackers from determining model behavior in the event of a breach, and should enforce strict governance over AI systems with close monitoring of data flows.
CSO OnlineData poisoning (corrupting training data to make AI systems behave incorrectly) has become much easier and more accessible than previously thought, requiring only about 250 poisoned documents or images instead of thousands to distort a large language model (an AI trained on massive amounts of text). Adversaries ranging from activists to criminals can now inject harmful data into public sources that feed AI training pipelines, and the resulting damage persists even after clean data is added later, making this a major security threat for any organization using public data to train or update AI systems.
Fix: One of the most reliable protections is establishing a clean, validated version of the model before deployment, which acts as a 'gold' version that teams can use as a baseline for anomaly checks and quickly restore to if the model starts producing unexpected outputs or shows signs of drift.
CSO OnlineKey management (the process of creating, storing, rotating, and retiring cryptographic keys throughout their lifetime) is often overlooked in organizations despite being critical to security, and this gap becomes even more dangerous as post-quantum cryptography (encryption designed to resist quantum computers) and AI systems become more widespread. The real challenge of post-quantum readiness is not choosing the right algorithm, but building operational ability to safely rotate and manage keys across systems without downtime. AI systems introduce additional risks because keys protect not just data access but also AI behavior and decisions, requiring tighter key controls and more frequent rotation than traditional applications need.