Claude 4 benchmarks show improvements, but context is still 200K
Today, OpenAI rival Anthropic announced Claude 4 models, which are significantly better than Claude 3 in benchmarks, but we're left disappointed with the same 200,000 context window limit.
This publication provides essential data security guidance for organisations that develop and/or use AI systems, including businesses, government and critical infrastructure. It highlights the importance of data security in ensuring the accuracy and integrity of AI outcomes, and presents an in-depth examination of 3 areas of data security risks in AI systems: data supply chain, maliciously modified (poisoned) data, and data drift.
The U.S. government today unsealed criminal charges against 16 individuals accused of operating and selling DanaBot, a prolific strain of information-stealing malware that has been sold on Russian cybercrime forums since 2018. The FBI says a newer version of DanaBot…
The alleged leader of the cybercriminal gang behind the Qakbot malware, which was used by many high-profile ransomware gangs, has been indicted by the U.S. Justice Department.
News Alert: Halo Security reaches SOC 2 milestone, validating its security controls and practices
Miami, Fla., May 22, 2025, CyberNewsWire -- Halo Security, a leading provider of attack surface management and penetration testing services, today announced it has successfully achieved SOC 2 Type 1 compliance following a comprehensive audit by Insight Assurance. This certification validates that Halo Security's security controls and practices are properly designed and implemented to meet
US indicts leader of Qakbot botnet linked to ransomware attacks
The U.S. government has indicted Russian national Rustam Rafailevich Gallyamov, the leader of the Qakbot botnet malware operation that compromised over 700,000 computers and enabled ransomware attacks.
News alert: INE Security, Abadnet Institute partner to deliver cybersecurity training in Saudi Arabia
Cary, NC. May 22, 2025, CyberNewswire -- INE Security, a global leader in Cybersecurity training and certifications, has announced a strategic partnership with Abadnet Institute for Training, a Riyadh-based leader in specialized Information Technology, Cybersecurity, and Networking training. The collaboration leverages INE Security's internationally recognized cybersecurity training content and Abadnet's established presence in the Saudi
Hazel observes that cybercriminals often fumble teamwork, with fragile alliances crumbling over missed messages. Plus, how UAT-6382 is exploiting Cityworks and what you can do to stay secure.
Cyberattaque de Marks & Spencer : plus de 350 millions d'euros de pertes
Le géant britannique de la distribution a été victime d'une cyberattaque en avril, entraînant notamment une suspension des commandes en...-Cybersécurité
Chinese-speaking hackers targeting US municipalities with Cityworks bug
Since January, cybersecurity experts have seen Chinese-speaking hackers exploiting a bug impacting a tool used by local governments to manage critical infrastructure assets and other services.
Hackers use fake Ledger apps to steal Mac users’ seed phrases
Cybercriminal campaigns are using fake Ledger apps to target macOS users and their digital assets by deploying malware that attempts to steal seed phrases that protect access to digital cryptocurrency wallets.
Police arrests 270 dark web vendors, buyers in global crackdown
Police arrested 270 suspects following an international law enforcement action codenamed 'Operation RapTor' that targeted dark web vendors and customers from ten countries.
Ivanti EPMM flaw exploited by Chinese hackers to breach govt agencies
Chinese hackers have been exploiting a remote code execution flaw in Ivanti Endpoint Manager Mobile (EPMM) to breach high-profile organizations worldwide.
[tl;dr sec] #280 - Hardening GitHub Actions, Uber's Multi-Cloud Secret Management, Prompts are the New IOCs
Detailed guide on all the ways to harden GitHub Actions, Uber's Secret Management Platform that manages >150,000 secrets, "LLM TTPs" + a tool to detect and hunt adversarial prompts