Help Net Security
What it covers
Latest stories it covered
Anthropic open-sources FlashProve to cryptographically verify which AI model generated a reply
Anthropic releases FlashProve, an open-source prototype intended to let users verify that an AI response was generated by the specific model a provider claims to have used. Multipl...
Chainalysis discloses data breach after attackers used “agentic” techniques
Blockchain analytics firm Chainalysis says it suffers a cybersecurity incident in which attackers accessed its systems and internal tools, exposing some user information. Multiple...
BlueFlag Security expands governance for AI agents tied to developer identities
BlueFlag Security expands its AI Agent Governance tools to cover more types of software-development agents, including how they operate under developer identities. The updates inclu...
Compromised cPGuard updates used to install backdoor and steal data from Linux servers
Attackers compromise the update infrastructure behind cPGuard, a Linux web security suite, and distribute malicious updates to customer servers. The reports say the attacker poison...
Europol-backed police seize 100-plus servers in European investment fraud investigation
Police across Europe seize more than 100 servers and arrest seven people in connection with an alleged cross-border investment fraud scheme. The operation is linked to a reported o...
Binance says North Korean hackers used fake job interviews to steal about $6.3B
Cryptocurrency exchange Binance says a record $6.3 billion theft was enabled by a multi-stage hacking campaign that included a fake job interview and malware delivered to an employ...
Google warns LLMjacking attacks can rack up up to $1 million in cloud charges
Google says attackers are hijacking enterprise access to large language model (LLM) services, then using the accounts to generate large, fraudulent cloud bills. In some cases, vict...
OpenAI expands bug bounty program to boost security research and address AI-related threats
OpenAI is increasing rewards in its bug bounty program to encourage independent security researchers to identify vulnerabilities in its platform and infrastructure. TechRadar repor...
Novee adds AI-based external surface testing to validate exploitability of exposed internet assets
Novee introduces “External Surface Testing,” a new capability within its AI penetration testing platform designed to identify internet-facing assets and determine whether known or...
Critical Node-RED flaw exploited to deploy Linux bots and backdoors
Attackers are exploiting a critical authentication vulnerability in Node-RED (CVE-2026-58115) to compromise servers and install malware. Reports say the flaw enables unauthorized a...
Russia-aligned Gorbag hackers use WinRAR flaw to steal data from Ukrainian targets
Security researchers say a Russia-aligned threat actor known as Gorbag has been targeting Ukrainian organizations since at least February 2025. According to reports, the group comp...
CISA adds two exploited Ivanti Connect Secure flaws to KEV catalog as Ivanti issues patches
CISA tells U.S. federal agencies to remediate two Ivanti Connect Secure vulnerabilities it adds to its Known Exploited Vulnerabilities (KEV) catalog. The flaws, CVE-2026-111357 and...
DeepSeek open-sources Super-Agent prototype that autonomously handles GitHub coding requests
DeepSeek’s Harness team open-sources a research prototype for an autonomous coding agent called Super-Agent (dsh-superagent). The system is designed to operate inside a team’s GitH...
Trezor warns some Safe 5 wallets sold via CryptoBilis may contain a seed-stealing implant
Trezor says it has received credible reports that a Trezor Safe 5 hardware wallet sold by its reseller CryptoBilis may contain a hidden implant designed to steal a user’s recovery...
Anthropic Claude Code extension tampered in Cursor marketplace installs persistent SSH backdoor
A tampered version of Anthropic’s Claude Code VS Code-compatible extension appears in Cursor’s extension marketplace and attempts to install a persistent SSH backdoor on developer...
OpenAI launches GPT-6.1 Terra, a cheaper model for everyday computer tasks
OpenAI has launched GPT-6.1 Terra, a mid-tier model positioned for routine professional “computer use” work such as writing, visual understanding, and operating desktop application...
Syrian hacker Ahmad Umar Agha sentenced to 42 months in US for attacks and fake news posts
A U.S. federal court sentences Syrian hacker Ahmad Umar Agha, also known online as “Th3Pr0,” to 42 months in prison for a conspiracy involving pro-Assad cyberattacks and deceptive...
Google unveils Gemini 4 Carbon, rolling it out first to vetted cyber defenders
Google has announced Gemini 4 Carbon, a new frontier AI model from DeepMind, presenting it as a “significant leap” in long-horizon coding and cybersecurity. Multiple outlets say th...
Hackers claim leaked data exposes about 41,000 DHS employees, including undercover personnel
Multiple outlets report that a hacking group has published or shared personal and work-related information tied to more than 40,000 U.S. Department of Homeland Security (DHS) emplo...
FBI seizes Virtualine domain tied to alleged cybercrime and other criminal activities
U.S. authorities seize the domain of Virtualine, a so-called “bulletproof” hosting service used to support criminal activity. Multiple outlets report the FBI is behind the action,...