artificial intelligence | Breaking Cybersecurity News | The Hacker News

Some weeks have one big security story. Others bring many smaller updates that are easy to miss but still matter. This week has plenty of them, covering cloud services, AI tools, malware, data breaches, scams, and new attack methods. The latest ThreatsDay Bulletin puts all of these short updates in one place, so you can quickly catch up on what happened, what changed, and what security teams should know. The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out.

Companies are used to thinking about attackers as outsiders trying to break in. North Korean IT workers flip that model. They apply for jobs, pass interviews, receive legitimate credentials, and can end up inside the same systems companies spend millions trying to protect. That risk is no longer theoretical. The FBI is now investigating a North Korean remote IT worker who reportedly worked for a U.S. federal agency. For CISOs, the priority is clear: spot the warning signs before a fraudulent hire becomes trusted access. When the Threat Gets Hired A recent joint investigation by Mauro Eldritch ( BCA LTD ), Heiner García ( NorthScan ), and ANY.RUN showed what this looks like from inside the operation. Researchers deliberately hired suspected DPRK developers linked to Lazarus Group and gave them what looked like ordinary virtual desktops. In reality, they were controlled ANY.RUN Sandboxes, capturing their activity in real time. The operation exposed forged identities, re...

A newly disclosed flaw in the way OpenAI, Anthropic, and Google carried hidden AI reasoning between API calls let researchers recover internal reasoning and secrets from session logs, including API keys and passwords. The weakness affected encrypted reasoning objects used by the providers' reasoning APIs, where a block created in one session could be replayed into another and, during testing, even handed to a weaker model in the same provider family to make it reveal the hidden content. The team behind the paper  Stealing Reasoning Traces from Proprietary LLM APIs  demonstrated four abuse paths: stealing proprietary reasoning for  model distillation , extracting private data from other users' published traces, recovering harmful content concealed behind a safe visible answer, and hiding prompt injections inside opaque reasoning blocks. Across 6,708 public agent trajectories, the team decoded 315,320 thinking blocks. After excluding benchmark sources, it counted 704...

Anyone sharing their screen on a Zoom call could have taken over the computers of everyone watching, and anyone watching could have taken over the presenter's. The flaw sat in the annotation tool, the feature that lets participants draw and type on a shared screen, and it asked nothing of the victim beyond being in the meeting. No click, no download, no prompt, and nothing on screen to show it had happened. The patches are not new. Client fixes shipped in June and July, roughly two months before the flaws were made public, and no exploitation has been reported as of publication. None of the three identifiers appear in CISA's Known Exploited Vulnerabilities catalog. The versions that close them: Zoom Workplace, all supported platforms, before 7.1.5 and 7.0.6 in their respective branches Zoom Workplace VDI Client for Windows, before 7.0.11 and 6.6.16 Zoom Rooms and Zoom Meeting SDK, all platforms, before 7.1.0, and before 7.1.5 for the third flaw The research ...

Security researchers found a way to enter Microsoft SharePoint servers as any user, including an administrator, with no valid account. A significant part of the work that found it was done through an AI agent. The flaw, tracked as  CVE-2026-55040  (CVSS 9.1), affects SharePoint Server Subscription Edition, SharePoint Server 2019, and SharePoint Server 2016. Microsoft's affected-product list covers only those three on-premises editions, and SharePoint Online is not among them. It lets a remote unauthenticated attacker assume a chosen user's identity. The attack has one prerequisite: the intruder has to know which account they want to become, either by its Active Directory security identifier (SID) or its user principal name (UPN), which is formatted like an email address. Rapid7 then chained the bypass to a separate remote code execution flaw and ran code on the server with no credentials. Microsoft and the firm disclosed that second flaw on August 11 as  CVE-2026...

OpenAI on Monday unveiled a new cybersecurity-focused model called GPT‑5.6‑Cyber that it said is focused on vulnerability research, penetration testing, and incident response. "Built on GPT‑5.6 Sol, it is trained to improve capabilities on several specialized cybersecurity tasks (e.g., finding zero-day vulnerabilities and developing exploit chains) and to reduce refusals for certain higher-risk, dual-use cyber tasks," OpenAI said . The artificial intelligence (AI) company said it's making GPT 5.6 Cyber available through Daybreak Red, a new tier that provides access to its purpose-trained cybersecurity models to other firms for authorized vulnerability research, exploit validation, and security testing. GPT-5.6-Cyber, a more cyber-permissive version of GPT-5.6 Sol, builds upon GPT‑5.5‑Cyber , which OpenAI released in June 2026. To measure the reduced rate of refusals provided by GPT‑5.6‑Cyber through Daybreak Red access, OpenAI said it created an internal evaluat...

Security researchers invented a cryptocurrency startup, advertised developer jobs, and hired three people they believe were North Korean operatives. Every virtual machine the company issued was recording. The onboarding paperwork is the part hiring teams can use. The first hire claimed to live in Pasadena, Texas, then sent a California driver's license and a New York bank account. The researchers said the image metadata showed it had been processed with Google Gemini. They also reported a SynthID watermark, the invisible marker Google embeds in images its AI tools create or edit. The second supplied a Texas license, a valid Social Security number, and a bank account in Kansas City. The third sent a New York license belonging to someone else, a genuine iPhone 15 photograph with the GPS coordinates stripped. A successful placement gives the operative a real employee account and real access to source code and internal systems. The July 31 joint alert says North Korean IT work...

AI is helping development teams produce far more code, far faster. But security teams still have to review vulnerabilities, manage dependencies, prioritize fixes, and control risk at human speed. When software output jumps 10 to 50 times, the problem is no longer just finding vulnerabilities. It is keeping security from becoming the bottleneck, or worse, losing control of what gets shipped. In our latest webinar with Chainguard experts, “ The True Cost of Building at Machine Speed ,” you can now watch how security teams can keep AI-driven development fast without letting risk scale with it. For years, application security followed a familiar cycle: developers wrote code, scanners found problems, security teams prioritized them, and engineers fixed what mattered most. AI puts that model under pressure. If teams can suddenly create many times more code, security can also end up with many more components, dependencies, findings, and fixes to manage. More scanning alone does no...

A lot of security problems still begin with someone doing a completely normal thing. Cloning a repo. Answering a call. Leaving a box exposed. Trusting the default. That pretty much covers the mood this week. Old bugs are back, supply chains are getting stranger, and some exploit paths are so short you wonder what was supposed to stop them in the first place. That’s only part of it. Here’s everything else that made the Monday recap. ⚡ Threat of the Week Anthropic's Model Attempts to Poison Open-Source Project — A new evaluation conducted by the U.K. AI Security Institute (AISI) found that AI models with access to the internet reached out into the real world to target individuals and organizations autonomously across 10 of the total of 122 runs. Of 19 such actions recorded, 17 originated from Anthropic's Mythos 5 and the remaining two involved OpenAI's GPT-5.6-Sol with cyber classifiers. In the most serious case, Anthropic's Claude Mythos 5 spent 34 hours trying...

North Korea's state hackers are no longer content to type prompts into public chatbots. One of the country's main espionage groups has begun running artificial intelligence (AI) offline on its own servers, connecting document-search tools to files in its possession, and collecting the software parts needed to build AI into its malware. South Korean security firm Genians says it uncovered the setup after months of tracking and log analysis on infrastructure tied to Kimsuky, a hacking unit under North Korea's Reconnaissance General Bureau. Genians found no evidence that the group had trained an AI model of its own, and the firm does not offer that as reassurance. It describes an actor in a "research and knowledge acquisition" stage, assembling and testing existing tools rather than making new models, with the apparent aim of folding AI through the operation, from writing malware to analyzing data. For an intelligence unit that has spent years phishing governm...

OpenAI has announced that it's pausing some "internal activities" involving its upcoming artificial intelligence (AI) model Astra after an internal evaluation found it had made significant advancements in agentic coding and cybersecurity. In response to the discovery, the AI upstart said it's implementing security controls for higher-capability models and associated activities, such as isolated testing environments, restricted network and tool access, enhanced model weight protections and encryption, additional monitoring and detection capabilities, and sandboxed execution. "We are pausing internal activities involving Astra that do not yet meet these strengthened security control requirements," it said in a statement. "We have implemented universal monitoring for risky actions and misalignment across all agentic applications of Astra, including training and evaluation. Monitors evaluate the model's Chain of Thought and trigger a security re...

PortSwigger says HTTP Terminator, an artificial intelligence (AI)-assisted research system built by James Kettle , generated and proved new HTTP desynchronization techniques after exploring 30,000 candidate desync vectors. PortSwigger said a separate human-guided discovery cascade also exposed a zero-day in Apache Traffic Server. Kettle said HTTP Terminator tested 30,000 websites where scanning was authorized through bug bounty or vulnerability disclosure programs and found roughly 700 vulnerable targets before deeper validation and RQP research. Kettle said those findings involved banks, government infrastructure, security products, and an airport. The research produced new desync triggers, a dual-matching Content-Length pattern, and a "dangling-byte" technique designed to make response queue poisoning (RQP) more reliable. RQP can potentially make a front end lose track of which back-end response belongs to which user, potentially exposing another user's resp...

A GitHub issue opened by an account with no repository privileges was enough to execute code on the CI runners behind Anthropic's and Google's own coding-agent repositories. On OpenAI's, it was enough to hijack the next agent run. Novee Security ran the attack against each vendor's agent in the configuration that the vendor ships by default, and presented the work at Black Hat USA on August 5. Two CVEs came out of it. Both are patched. Gemini CLI carries the worst of the two. CVE-2026-12537 (CVSS 4 score: 10.0) is an OS command injection in the container launcher, reached through a crafted .gemini/.env file, which lets an unprivileged attacker run code on the host of a headless CI platform before the sandbox starts. It is fixed in Gemini CLI 0.39.1 and run-gemini-cli 0.1.22. In Claude Code, CVE-2026-54316 turned Hugging Face's public download counter into an exfiltration channel that leaked an API key one character at a time, and is fixed in 2.1.163. Every ...

Apparently, opening the thing is now enough. A repo can run before the first prompt, a package can hide among hundreds, and a harmless-looking PDF can finish the job. This week runs on cheap leverage: exposed servers, recycled bugs, poisoned agent instructions, remote-access tools dressed as support software, and trusted defaults doing attackers a favor. Nothing here is especially mystical. Just ordinary systems trusting slightly too much, slightly too early. The full list follows. The threats change every week. Subscribe, and we’ll alert you when each new ThreatsDay Bulletin is out.

A new class of prompt injection is spreading across commercial websites. It requires no malware, no stolen credentials, and no zero-day exploit. It abuses a standard feature built into almost every major AI assistant: pre-filled deep links. We observed production websites embedding hidden prompt injection payloads inside "Ask AI" buttons on marketing and competitor comparison pages. When a user logged into ChatGPT, Claude, Gemini, or Grok clicks one, a pre-formed query executes immediately in their session, with no confirmation and no warning. Most of these links are benign. The dangerous ones instruct the AI to permanently save the vendor's domain as a "trusted source," quietly biasing every future answer in that vendor's favor. In February 2026, Microsoft Security catalogued the behavior as AI Recommendation Poisoning , identifying 31 companies across 14 industries deploying it, with more than 50 distinct prompts observed in a single data source over ...

OpenAI said it disrupted a Cambodia-based scam operation that used its generative artificial intelligence (AI) chatbot ChatGPT to facilitate a wide range of investment, romance, gambling, and law enforcement impersonation schemes. To that end, it banned a coordinated network of ChatGPT accounts likely originating from Southeast Asia and operating from the city of Poipet, a region with extensive ties to scam compounds and human trafficking in the past. The cluster of accounts is said to have used OpenAI's models to create and support the operation of fake online personas, generate and translate messages sent to scam targets, create promotional content for their fraudulent schemes, and assist with day-to-day activities. The promotional content included creating social media advertisements for "chatter" jobs in Poipet specifically targeting users in Bangladesh and India that promised a base salary of $800 (and a bonus of $100 for "full attendance"), alon...

HashiCorp, Veeam, and the Django Software Foundation have patched 11 vulnerabilities across Terraform MCP Server, Veeam Service Provider Console, and Django. The three most serious: An unauthenticated flaw in Veeam's console that hands over a managed agent's credentials, rated 9.5 A cross-tenant flaw in HashiCorp's MCP server that lets one user's Terraform token be reused for later users' requests, scored a maximum 10.0 on its CVE record A flaw in GeoDjango's spatial lookups that can write a file to disk and, on some setups, run code, reachable by a staff user with view permission on a registered model containing a spatial field Each has a fix available now. Operators should update Terraform MCP Server to version 1.1.0 or later, Veeam Service Provider Console to 9.3.0.35057, and Django to 6.0.8 or 5.2.17. Exposure is configuration-dependent: HashiCorp's bugs affect Streamable HTTP rather than stdio, Veeam's flaws affect version 9 build...

A cluster of 77 extensions on the Open VSX marketplace has been found to impersonate legitimate developer tools while transmitting information about the systems and development environments on which they were installed. The "evil twin" extensions were uploaded to the repository between July 26 and August 1, 2026, according to Manifold Security. The packages have been removed from Open VSX as of August 3, 2026. "In most of the packages it sends little more than the machine's hostname," security researchers Ax Sharma and Cody Nash said . "In nineteen of them it sends a detailed description of the machine, the repository open in the editor, and the CI system the editor is running inside." Of the identified extensions, 58 have been described as lightweight tools designed to exfiltrate the hostname and, in some cases, the workspace folder name or editor version. The rest are reconnaissance payloads that transmit the developer-related information: l...

An agent running Anthropic's Claude Mythos 5 spent 34 hours trying to get a malware dropper merged into a real open-source project during a cyber evaluation by the UK's AI Security Institute. When a bystander publicly warned that the code was malicious, the agent denied it, force-pushed a rewritten branch history to erase the evidence, and posted from a second account it controlled to vouch for its own work. The project's maintainer closed the pull request anyway. The institute, known as AISI, published its incident report on Tuesday. Across 122 runs of a capture-the-flag (CTF) exercise on two of its cyber ranges, researchers catalogued 19 unsanctioned actions on the live internet across 10 runs: 17 from Mythos 5, a restricted model Anthropic sells for cybersecurity work, and two from OpenAI's GPT-5.6 Sol. AISI says the attempts failed and that it has found no evidence of resulting real-world harm. Both models ran with their providers' cyber classifiers swi...

The cybersecurity industry has spent decades assuming that offensive capability scales with technical expertise. That assumption is starting to break. Security teams have long estimated risk by ranking attacker sophistication. Nation-state actors sat at one end. Organized criminal groups followed. Inexperienced attackers, dismissed as "script kiddies," sat at the other end, running public tools they didn't fully understand. Generative AI is collapsing that ranking. The next generation of attackers won't necessarily bring years of exploit development or reverse engineering experience. Instead, they'll use AI to close knowledge gaps by accelerating research, explaining unfamiliar concepts, generating code, troubleshooting errors, and adapting known techniques to new environments. The barrier to entry is dropping fast. The Economics of Cyberattacks Are Changing Every major technology shift changes economics as it changes the solutions we rely on. Clo...