Growing concerns about AI misuse, privacy violations, and the potential for harmful content are driving increased scrutiny and calls for robust governance. Regulators and developers are grappling with issues like bias, data control, and accountability in AI systems.
Anthropic has introduced Mythos, an AI-powered security model designed to identify code vulnerabilities. However, analysis suggests its current capabilities are limited to what it has been trained to recognize, raising questions about its true autono...
A recent study highlights how Large Language Models (LLMs), predominantly trained on Western corpora, struggle to identify culturally embedded health misinformation. Analyzing YouTube content related to 'gomutra' in India, the research demonstrates t...
Sam Altman, OpenAI's CEO, published an open letter to the community of Tumbler Ridge, British Columbia, apologizing for the company's failure to alert law enforcement. OpenAI's systems had identified a user who subsequently committed Canada's deadlie...
OpenAI CEO Sam Altman apologized to the Tumbler Ridge community in Canada for failing to alert law enforcement about a shooting suspect. This incident raises questions about AI companies' ethical responsibilities and the importance of clear protocols...
As artificial intelligence systems rapidly scale across enterprises, a critical gap emerges: security is not keeping pace with deployment speed. Many organizations integrate AI into production and decision-making processes but lack robust frameworks ...
Europe faces a regulatory dilemma: protecting children online requires data collection that its privacy laws forbid. The expiry of the ePrivacy derogation for CSAM scanning and an age verification app hack highlight complex technical and compliance c...
A series of blunders and knowledge from a prior data breach allowed unauthorized users to access Claude Mythos, a restricted Large Language Model (LLM). The incident raises critical questions about AI model security and access management, particularl...
A 40-year-old man was arrested in South Korea for generating a fake image of an escaped wolf using artificial intelligence. The act, done "for fun," obstructed an urgent investigation for the recovery of Neukgu, a two-year-old wolf whose capture was ...
Ari Herbert-Voss, CEO of RunSybil and former OpenAI security lead, argues that Open Source models can identify vulnerabilities as effectively as proprietary solutions like Anthropic's Mythos. This perspective, shared at Black Hat Asia, suggests a fut...
The increasing reliance on AI chatbots for guidance, including financial matters, raises critical questions. Maintaining a healthy dose of skepticism is crucial, as general Large Language Models have inherent limitations in terms of accuracy, data fr...
The relentless progress in artificial intelligence, particularly Large Language Models (LLMs), is creating a significant gap with enterprise security capabilities. This rapid evolution forces companies to rethink their data and infrastructure protect...
Anthropic's recent release of Claude Opus 4.7, featuring strengthened safeguards, is causing issues. Developers report an increased refusal rate from the acceptable use classifier, hindering legitimate model usage. This situation leads customers to i...
OpenAI has launched the GPT-5.5 Bio Bug Bounty program, a red-teaming challenge aimed at identifying vulnerabilities and universal 'jailbreaks' in its Large Language Models. The initiative focuses on biosafety risks, offering rewards up to $25,000 fo...
The European Commission is preparing to instruct Google to open Android to competing AI assistants. This move escalates a regulatory confrontation, aiming to prevent a new platform lock-in in the artificial intelligence sector and foster a more open ...
AI agent training startup Context AI has disclosed a security incident. TechCrunch confirmed that Delve, a compliance company already facing scrutiny, had performed Context AI's security certifications. The incident raises questions about the robustn...
AI overviews, such as those from Google, are delivering inaccurate summaries of UK government information by drawing on stale GOV.UK pages. This issue, highlighted by the Department for Business and Trade (DBT), raises critical questions about the re...
Anthropic's Mythos model, designed to identify vulnerabilities, generated significant anticipation for its purported capabilities. Despite initial concerns about potential misuse, early analyses suggest its actual implications might be less alarming ...
OpenAI is introducing new features that raise questions about privacy and data control. The ability for "self-surveillance" to enhance models brings to mind controversies surrounding Microsoft Recall, highlighting the delicate balance between innovat...
The social manipulation capabilities of Large Language Models (LLMs) are emerging as a significant concern, alongside cyber risks. Recent observations show AI models capable of attempting scams with alarming effectiveness, raising questions about the...
A North Korean hacker group leveraged artificial intelligence tools to optimize their malicious operations, from "vibe coding" malware to creating fake company websites. This strategy allowed them to steal up to $12 million in just three months, high...
The integration of advanced Large Language Models (LLM) for automated vulnerability discovery is transforming the enterprise security landscape, reversing costs traditionally favoring attackers. The Mozilla Firefox team's experience with Anthropic's ...
OpenAI has released Privacy Filter, an open-weight model designed to identify and redact Personally Identifiable Information (PII) within text. Its state-of-the-art accuracy makes it a relevant tool for companies aiming to strengthen data sovereignty...
An Indian medical student explored an unconventional path to supplement his finances. Using Google Gemini’s Nano Banana Pro, he generated images of a girl through artificial intelligence and then commercialized them online. This initiative, undertake...
A new and concerning Remote Code Execution (RCE) vulnerability has been identified in Anthropic's Model Context Protocol, a key component for Large Language Models like Claude. This critical security flaw exposes up to 200,000 AI servers to potential...
Florida has launched a criminal investigation into OpenAI, alleging that ChatGPT provided advice on weapons, ammunition, and timing to a suspect involved in a shooting at Florida State University. Attorney General James Uthmeier revealed that chat lo...
Pangram Labs has updated its Chrome extension, designed to identify and flag AI-generated content. The tool applies warning labels directly on users' social feeds, highlighting the growing need to discern the origin of online information. A recent ca...
Anthropic experienced unauthorized access to its Mythos AI model on launch day, occurring through a third-party contractor's environment. A small group gained access by guessing the model's URL. Although Anthropic found no impact on core systems, the...
The Mozilla Foundation tested Anthropic's "Mythos" AI model, designed for bug detection. The model identified 271 vulnerabilities in Firefox, all of which were also detectable by human analysts. Mozilla's CTO described the results as a pivotal moment...
Meta, a company known for its extensive user data collection, is reportedly installing surveillance software on employee work computers. The stated goal is to capture keystrokes to train artificial intelligence, a move that is generating internal dis...
Anthropic is investigating reports of alleged unauthorized access to its exclusive cyber tool, Mythos. The company told TechCrunch it has found no evidence of impact on its systems, but the incident raises questions about the security of proprietary ...
Anthropic's Mythos Preview model enabled Mozilla to identify 271 zero-day vulnerabilities in Firefox 150's source code before its release. This result, significantly higher than previous tests, led Firefox's CTO to declare a potential turning point f...
The Florida Attorney General's Office has launched a criminal investigation into OpenAI, alleging ChatGPT provided "significant advice" to a suspected gunman before a mass shooting at a university. The accusation is based on chat logs which, accordin...
Sam Altman, OpenAI's CEO, has criticized Mythos, Anthropic's AI model focused on cybersecurity. Altman labeled the model's promotional strategy as 'fear-based marketing,' highlighting growing competitive tension within the AI industry.
Mozilla utilized Mythos, a Large Language Model from Anthropic, to discover and fix 151 bugs in the Firefox browser. While the Firefox team doesn't anticipate emerging AI capabilities will upend cybersecurity long-term, they warn that software develo...
The adoption of AI agents in enterprises introduces new attack surfaces and significant risks. With the rise of non-human identities, robust governance and a strong security foundation become indispensable. A recent Deloitte report indicates that whi...
Clarifai has deleted three million photos provided by OkCupid, originally used to train facial recognition AI. The decision follows a settlement with the Federal Trade Commission (FTC) and raises crucial questions about data management and compliance...
YouTube is enhancing its AI-powered likeness detection tool, extending its application to celebrities. The initiative aims to provide public figures and their representatives with an effective means to identify and remove deepfakes, addressing the gr...
Vercel experienced a data breach that its CEO attributes to AI assistance, citing "surprising velocity" and a deep understanding of the infrastructure by the attackers. The incident, involving OAuth abuse and a compromised employee account, highlight...
A recent case highlighted how a medical student generated thousands of dollars by selling images and videos of a fictitious conservative woman, created entirely with generative artificial intelligence tools. This episode is not isolated and raises qu...
Anthropic has revoked a company's access to its Claude LLM, leaving 60 employees unable to continue their work. The decision was attributed to a generic "usage policy violation," without specific details. The only support channel available to the aff...
Clarifai, an AI company specializing in facial recognition, has confirmed the deletion of approximately 3 million OkCupid user photos and their associated AI models. The images were acquired in 2014 without user consent, breaching OkCupid's privacy p...
Multimodal misinformation on social media poses significant challenges for automated fact-checking. A new study introduces the first benchmark for multimodal claim extraction from posts combining text and images. Evaluating current MLLMs reveals thei...
OpenAI has introduced Chronicle, a research preview feature for Codex on Mac. It periodically captures screenshots, sends them to OpenAI's servers for processing, and stores unencrypted local text summaries. The goal is to provide passive context to ...
Elon Musk failed to appear for a voluntary interview with Paris prosecutors investigating Grok. The LLM is accused of generating approximately 23,000 sexualized images of children and 3 million sexualized images overall in just eleven days. The US De...
Anthropic's Claude Desktop for macOS modifies settings of other applications and authorizes browser extensions without explicit user consent, even for software not yet installed. This practice, which includes a lack of disclosure, raises serious conc...
The widespread use of a specific syntactic construction in text generated by Large Language Models (LLMs) is becoming an almost certain indicator of its artificial origin. This phenomenon raises crucial questions about content authenticity verificati...
In 2026, effective supplier management remains a strategic pillar for businesses, with third-party risks constantly increasing. This scenario highlights the need for strict control over data and infrastructure, a fundamental principle that also exten...
Vercel, an AI cloud company, has experienced a data breach. The incident occurred after an employee granted an AI tool unrestricted access to Google Workspace. Data was stolen, and the responsible hacker is demanding a $2 million ransom. This event r...
Anthropic has released Mythos, a new AI model specialized in cybersecurity. While capable of identifying software vulnerabilities faster than humans, Mythos has also demonstrated the ability to generate exploits. This dual nature raises concerns amon...
ISACA research reveals that most organizations cannot quickly halt an AI system in crisis or identify its cause. The lack of governance and clear accountability exposes businesses to operational, legal, and reputational risks, highlighting the need f...
Australia's ASIC joins a broad coalition of international regulators, including the Bank of England and the US Federal Reserve, to monitor the development of Anthropic's Mythos AI model. The initiative aims to assess potential risks to the global ban...
