OpenAI's new Guardrails framework, intended to enhance AI safety, has been quickly bypassed by researchers using prompt injection techniques. HiddenLayer experts showed attackers can exploit both the content-generating model and its safety assessor. This highlights persistent difficulties in safeguarding AI systems, as LLMs used for both generation and evaluation are susceptible to identical manipulations.
Microsoft has introduced Project Ire, a pioneering autonomous AI agent designed to detect and classify malware at scale. The system automates reverse engineering, demonstrating high precision in tests, and aims to enhance cybersecurity by reducing the need for human intervention and speeding up threat response.
Google has rolled out Gemini 1.5 Pro to Gemini Advanced subscribers, offering a massive 1 million token context window for advanced data analysis. The model supports multiple modalities, including video, audio, and large documents, and is available in over 35 languages.
OpenAI is reportedly developing a new, budget-friendly subscription called "ChatGPT Go," potentially priced under $20. This move could broaden access to advanced AI models, positioning the company more competitively in the evolving artificial intelligence market.
ChatGPT shared conversations are being indexed by major search engines, effectively turning private exchanges into publicly discoverable content accessible to millions of users worldwide. The issue first came to light through investigative reporting by Fast Company, which revealed that nearly 4,500 ChatGPT conversations were appearing in Google search results.
A new malware named LAMEHUG uses large language models to generate commands and steal data from Windows systems. Ukrainian authorities link it to the Russian hacking group APT28, signaling a new era in AI-driven cyber threats.
Researchers have developed White-Basilisk, a compact and efficient AI model for detecting software vulnerabilities. With just 200 million parameters, this model outperforms much larger systems by analyzing extensive codebases with unprecedented speed and a reduced energy footprint, challenging the 'bigger is better' philosophy in AI.
Google is implementing an AI model to estimate user ages, aiming to protect minors from inappropriate content. The system analyzes online activity to automatically apply safety measures like SafeSearch and YouTube restrictions, enhancing online safety for younger users across its platforms.
Google has introduced "Video Overviews" to its AI research assistant, NotebookLM. The feature automatically creates narrated video slideshows from a user's source materials, like documents and images, to help visualize and understand complex information. The feature is initially available in English.
Microsoft has introduced Copilot Mode, a new experimental feature for its Edge browser. This mode transforms Edge into an AI-powered browser, capable of managing tabs, answering queries, and performing tasks using voice commands, aiming to enhance user productivity and streamline web navigation.
Google has rolled out AI Mode, its advanced artificial intelligence search experience, in the United Kingdom. This feature, powered by the Gemini 2.5 model, provides conversational-style answers to complex queries and supports text, voice, and image-based inputs, marking a significant shift in search technology.
Orange Group disclosed a cyberattack on its information systems, causing service disruptions. Teams are working to restore services and have filed a complaint. No customer data exfiltration has been detected.
CISA has issued an urgent advisory regarding critical injection vulnerabilities in Cisco's Identity Services Engine (ISE), actively exploited by threat actors. These flaws, identified as CVE-2025-20281 and CVE-2025-20337, allow remote code execution with root privileges, posing a severe threat to network security.
The Tea app data breach has grown into an even larger leak, with the stolen data now shared on hacking forums and a second database discovered that allegedly contains 1.1 million private messages exchanged between the app's members.
OpenAI has launched "Shopping," a new ChatGPT feature enabling users to directly locate and purchase products via the chatbot interface. This strategic addition underscores the ambition of AI firms like OpenAI and Perplexity to become comprehensive "everything companies."
CIA Intelligence Website Breached, Exposing Sensitive Contracts and Personal Data
OpenAI is rolling out the ChatGPT Agent to Plus, Pro, and Team subscribers. This new tool transforms ChatGPT into an autonomous assistant that can perform complex tasks like web browsing, data analysis, and document creation by using its own virtual computer, all while keeping the user in control.
Insurance company Allianz Life has confirmed that the personal information for the "majority" of its 1.4 million customers was exposed in a data breach that occurred earlier this month.
A global outage is affecting the Microsoft 365 Admin Center, preventing administrators from accessing crucial management tools. The disruption, which began on July 24, 2025, is the second this week, raising concerns about service dependability and impacting IT operations worldwide. Microsoft is investigating the issue, primarily linked to underperforming US East region infrastructure, with no cyberattack indications.
A new ransomware strain named Gunra is targeting Windows devices globally using double-extortion tactics. It encrypts files, steals sensitive data, and demands ransom payments via the Tor network, impacting sectors like healthcare, pharmaceuticals, and manufacturing.
A new ransomware-as-a-service (RaaS) group named Chaos has emerged, targeting a wide variety of business verticals with double-extortion attacks. The group, believed to have ties to the former BlackSuit/Royal gang, employs sophisticated techniques to maximize impact and hinder recovery.
The U.S. Department of the Treasury's Office of Foreign Assets Control (OFAC) has sanctioned three North Korean nationals and a company for supporting fraudulent IT worker schemes that generated illicit revenue for the Democratic People's Republic of Korea (DPRK) government.
Microsoft has removed a compatibility hold that prevented some Easy Anti-Cheat users from installing the Windows 11 2024 Update because of a known issue that triggers restarts with blue screen of death (BSOD) errors.
Hackers exploit Sharepoint 0-day to deploy Warlock ransomware, targeting servers and stealing cryptographic keys.
A hacker injected malicious wiping commands into version 1.84.0 of the Amazon Q extension for Visual Studio Code. Amazon removed the vulnerable version and released an update, stating that no customer resources were impacted.
White House AI and Crypto Czar, David Sacks, laid out the Trump administration's strategy for ensuring the United States maintains its global leadership in artificial intelligence. Sacks emphasized the critical importance of staying ahead of competitors like China, comparing the current AI landscape to the space race of the 20th century.
Hackers compromised Toptal's GitHub organization account and used their access to publish ten malicious packages on the Node Package Manager (NPM) index. Security researchers discovered the malicious packages and notified Toptal, which quickly removed the packages from NPM. Toptal is now investigating the incident to determine how the hackers gained access to their GitHub account and to prevent similar incidents in the future.
The US National Nuclear Security Administration was breached via a vulnerability in Microsoft SharePoint. Attackers were unable to access sensitive information. Microsoft blames Chinese hackers for the attack.
Microsoft has attributed recent attacks on SharePoint servers to Chinese threat groups, including Linen Typhoon, Violet Typhoon, and Storm-2603. Emergency patches have been released to address the vulnerabilities, and CISA has issued mitigation guidance for affected organizations.
Microsoft is rolling out significant changes to Windows 11 24H2 as part of the Windows Resilience Initiative. This initiative is designed to reduce downtime and help devices recover from serious failures. The update also includes an overhaul of the all-too-familiar BSOD crash screens.
A new variant of the banking trojan 'Coyote' has begun abusing a Windows accessibility feature, Microsoft's UI Automation framework, to identify which banking and cryptocurrency exchange sites are accessed on the device for potential credential theft.
CISA and the FBI warned on Tuesday of increased Interlock ransomware activity targeting businesses and critical infrastructure organizations in double extortion attacks.
AMEOS Group, an operator of a massive healthcare network in Central Europe, has announced it has suffered a security breach that may have exposed customer, employee, and partner information. The security breach at AMEOS Group is a significant incident, given the large volume of sensitive data the organization handles. The implications could be far-reaching, affecting many individuals and organizations. Authorities are investigating the incident.
The World Leaks group accessed and released data from the company's Customer Solution Center. This center is separated from customer and partner systems. It primarily stores "synthetic" datasets used for demos and testing, Dell said.
The Clear Linux OS team has announced the shutdown of the project, marking the end of its 10-year existence in the open-source ecosystem. This decision marks a shift in Intel's strategy and its focus on other areas. The Clear Linux OS team thanked the community for its support and contributions over the years.
Ring is warning that a backend update bug is responsible for customers seeing a surge in unauthorized devices logged into their account on May 28th. The company stated that the issue was not caused by a system breach or credential exposure. A Ring representative stated that the company is investigating the issue and working to resolve it. They encouraged users to enable two-factor authentication to protect their accounts.
ExpressVPN has fixed a flaw in its Windows client that caused Remote Desktop Protocol (RDP) traffic to bypass the virtual private network (VPN) tunnel, exposing the users' real IP addresses.
Attackers exploited a CrushFTP vulnerability to gain administrative access.
On July 19, 2025, Microsoft released an out-of-bound advisory addressing two vulnerabilities, one of which being rated as critical and allowing unauthenticated remote attacker to execute arbitrary code on vulnerable systems. This critical flaw is actively being exploited in the wild. It is recommended isolating vulnerable system from the Internet, but also from internal systems, and running a compromise assessment before updating.
Microsoft SharePoint servers under attack via zero-day vulnerability. No patch available.
Artificial intelligence (AI) is evolving and can analyze large datasets and identify patterns, providing early warnings for financial crises. This enables proactive interventions and enhances risk management, but also requires strong governance for the responsible development of the technology.
The Japanese police have released a Phobos and 8-Base ransomware decryptor that lets victims recover their files for free, with BleepingComputer confirming that it successfully decrypts files.
A critical Citrix NetScaler vulnerability, tracked as CVE-2025-5777 and dubbed "CitrixBleed 2," was actively exploited nearly two weeks before proof-of-concept (PoC) exploits were made public, despite Citrix stating that there was no evidence of attacks.
To quash speculation of a cyberattack or BGP hijack incident causing the recent 1.1.1.1 Resolver service outage, Cloudflare explains in a post mortem that the incident was caused by an internal misconfiguration.
A threat actor has been deploying a previously unseen malware called OVERSTEP that modifies the boot process of fully-patched but no longer supported SonicWall Secure Mobile Access appliances. OVERSTEP is a rootkit that allows attackers to maintain access and control of compromised devices. This is achieved even after reboots or firmware updates.
An international law enforcement operation dubbed "Operation Eastwood" has targeted the infrastructure of the pro-Russian hacktivist group NoName057(16), responsible for distributed denial-of-service (DDoS) attacks across Europe and the US.
Google patched a Chrome zero-day vulnerability (CVE-2025-6558) exploited by attackers.
Google patched six Chrome security issues, including an actively exploited vulnerability.
Google rescued SQLite from a critical vulnerability.
OpenAI is following Perplexity and is working on its own AI-powered browser codenamed "Aura."
Google Gemini for Workspace has a security flaw. It can generate email summaries with malicious instructions. These summaries direct users to phishing sites without attachments or links.
ISACA has introduced the CISA Associate designation to assist individuals who have passed the CISA exam but lack the required experience. The new designation aims to bridge the skills gap and help students kickstart their careers in the IT audit field.
The new ransomware strain's aggressive multithreading and cross-platform capabilities make it a potent threat to enterprise environments.
Banks must adopt a skills-based approach to workforce planning to stay competitive and agile. AI is reshaping the industry, making it imperative for financial institutions to become truly AI-driven, while addressing talent shortages and challenges in attracting and retaining the right people.
ChatGPT is testing a new feature called 'Study Together' aimed at enhancing interactive learning. This feature offers guided questions and adaptive feedback, making it ideal for students and educators.
The European Union has launched the EuroQCI initiative to build a pan-European, quantum-secure communication infrastructure. The plan combines terrestrial and satellite networks to protect critical data from future quantum threats, aiming for full implementation by 2030.
Kostas Xiradakis, former VP of Product & Growth at Viva.com, has joined Snappi, the first Greek neobank, as its new Chief Commercial Officer. He will lead the marketing, sales, and customer experience departments to drive the company's growth and customer-focused innovation.
Cloudflare, which powers about 20% of the internet, now blocks AI crawlers by default. This move gives website owners control over their content, introducing a permission-based model and a potential "Pay Per Crawl" system to compensate creators for their work.
The International Criminal Court (ICC) in The Hague confirmed it sustained a "sophisticated and targeted" cyberattack. The incident, which was detected and contained, marks the second major attack against the court in recent years, raising security concerns amid high-profile investigations.
Two years after its initial discovery, a critical vulnerability in Microsoft's Entra ID, known as nOAuth, continues to expose thousands of SaaS applications to potential account takeovers. New research from Semperis reveals the flaw, which bypasses standard defenses, remains a severe and active threat.
ClickFix attacks surged over 500% in the first half of 2025, becoming the second most prevalent threat after phishing. ESET's report reveals ClickFix was responsible for nearly 8% of blocked attacks, using fake reCAPTCHA checks to trick victims.
The Ministry of Justice (MoJ) confirmed a significant data breach at the Legal Aid Agency (LAA) last month. This incident allowed criminals to access sensitive information belonging to hundreds of thousands of individuals, with records dating back to 2010. The compromised data includes personal contact details, dates of birth, national ID numbers, and financial information, posing a serious threat to those affected.
The EU's DORA regulation came into force on January 17, 2025, imposing strict requirements for the digital operational resilience of financial institutions. Businesses must manage ICT risks, report incidents, and test their resilience, ensuring the stability of the financial system.
