artificial intelligence | Breaking Cybersecurity News | The Hacker News

Introduction The average enterprise security team has 40 or more security tools, giving a lot of visibility into internal telemetry and asset data. But often, these tools are working in siloes, generating (overlapping) alerts and data. And yet, breach dwell times remain stubbornly long (~43 days), response windows keep closing before teams can act, and analysts burn out triaging noise instead of stopping threats. The problem isn't effort. It's architecture. Security programs were built for a world where threats moved slowly enough for humans to coordinate responses manually. That world no longer exists. With the way AI capabilities are getting developed and used, especially with frontier AI tools, a much more proactive stance to security is needed as well as machine speed response to combat fast moving adversaries. Gartner's Continuous Threat Exposure Management (CTEM) framework helps this shift from reactive, point-in-time assessments to a continuous, iterative cycl...

If an autonomous AI agent interacts with your company's core intellectual property today, can your security team instantly name the person who authorized it? For most enterprises, the answer is a simple no . The rush to adopt internal AI tools has left a massive trail of administrative debt: orphaned agents (AI tools left running after their creator leaves the company) and standing privileges (AI that retains permanent, unrestricted access it no longer needs). When an employee moves on, the automated tools they built stay active—often keeping unmonitored access to sensitive databases and source code long after the human’s credentials are revoked. To help security teams bridge this line of accountability, The Hacker News is hosting a technical briefing. Secure your spot today for the live webinar: Orphaned Agents & Standing Privileges: The Hidden Access Risks of Internal AI . Why Existing Security Tools Miss the Signal Traditional access tools treat AI like stand...

The internet did not break this week. It got used exactly as designed, which is worse. Searches were siphoned through shady browser add-ons. AI chat links turned into malware delivery paths. macOS attacks ran in memory and left almost nothing behind. Cloud agents looked like helpers until attackers treated them like open shells. Add exposed edge gear, poisoned packages, cash courier scams, stealers, loaders, and phishing that barely bothers pretending anymore. Here’s the full mess.

A default low-privilege account on a LiteLLM proxy can climb to full admin and run code on the server by chaining three vulnerabilities, researchers at Obsidian Security disclosed LiteLLM is a widely deployed open-source AI gateway that brokers calls to more than 100 model providers behind one OpenAI-compatible interface. A server takeover exposes every provider key it holds, the secrets that decrypt its stored credentials, and every prompt and response passing through it. Obsidian rates the full chain CVSS 9.9, in the Critical range. BerriAI , the maintainer, included the complete fix set in LiteLLM v1.83.14-stable, which GitHub lists as released May 2. Upgrade to that release or later to close the three-CVE chain. The three bugs The first link is CVE-2026-47101 , an authorization bypass. When a regular user (an internal_user) generates a virtual API key, LiteLLM stores the caller-supplied allowed_routes field without checking it against the user's role. The field is...

Stuff broke again. Not in a movie way. An old tool was left exposed. An abandoned package was abused. A deprecated feature was still running in prod. This week is the same lesson in a new form: phishing kits are easier to rent, AI names are useful bait, old login paths still fail, and forgotten software keeps becoming someone else's entry point. Scroll through the full Monday Cybersecurity Recap below for the news, tools, webinars, and fixes worth your time this week. ⚡ Threat of the Week Google Patches Actively Exploited Chrome 0-Day - Google released security updates to address 74 vulnerabilities, including one that has come under active exploitation in the wild. The high-severity vulnerability, tracked as CVE-2026-11645 (CVSS score: 8.8), has been described as an out-of-bounds memory access in V8, Chrome's JavaScript and WebAssembly engine. Google acknowledged that an "exploit for CVE-2026-11645 exists in the wild," but stopped short of sharing addition...

Anthropic said on Friday it will "abruptly disable" its most advanced artificial intelligence (AI) models, Claude Fable 5 and Mythos 5 , for all users after the U.S. government ordered it to suspend access to the models for foreign nationals, whether inside or outside the U.S., citing national security concerns. The AI company said it received an order at 5:21 p.m. ET, instructing it to suspend all access to the models by foreign nationals. It said that it believed there was a "misunderstanding" and that it is working to restore access to the models as soon as possible. Access to other models will not be affected by the export control directive. "Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5," the company said. "We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulner...

Google on Friday said it's pursuing legal action against a Chinese cybercrime network, accusing it of using its Gemini artificial intelligence (AI) agent to send phishing text messages targeting Americans. The network is said to be behind the development and management of a phishing-as-a-service (PhaaS) software kit called Outsider, per the tech giant. "The operation weaponized Gemini to help generate fraudulent phishing pages and deploy massive SMS phishing ('smishing') attacks, often through text messages impersonating legitimate brands, alerting recipients of 'brokerage account issues' or insisting they are eligible for 'rewards through their mobile phone carrier,'" Google said . "The texts prompt users to click a link leading to a fraudulent website that mimics trusted institutions to steal personal and financial information." Google said it's filing the lawsuit to dismantle the network's infrastructure, and that it...

Cybersecurity researchers have described what they say is a new class of attack that can trick artificial intelligence (AI) coding agents into running arbitrary code on developer machines. Called Agentjacking by Tenet Security, the attack can be triggered by means of a fake error report crafted using Sentry, an open-source error-tracking and performance-monitoring platform. "The attack exploits a critical architectural flaw at the intersection of Sentry's event ingestion (which accepts arbitrary payloads from anyone with the DSN) and the Sentry MCP server (which returns this data to AI agents as trusted system output)," security researchers Ron Bobrov, Barak Sternberg, and Nevo Poran said . The idea is to inject crafted input into Sentry error events, which are then interpreted by coding agents like Claude Code and Cursor as legitimate diagnostic resolution steps and run attacker-controlled code. A successful attack of this kind can expose sensitive data, includ...

It's been one of those weeks. You expect the usual noise: recycled malware, sloppy attacks, another easy target getting hit. Instead, there's a supply chain attack kit in a public repo, a $5,000-a-month RAT that clones browsers, and research showing AI agents can be tricked into leaking real credentials. The bigger problem is how polished this all looks now. Mule networks run like SaaS. Deepfake KYC bypass is sold as a feature. Endpoint tools can be quietly weakened using built-in OS settings, with no exploit needed. Here's the full list of threats, tools, flaws, and updates worth knowing.

For thirty years, vulnerability management ran on a buffer: the months between when a vulnerability was found and when someone could figure out how to weaponize it. The solution was straightforward enough; triage by severity, schedule the fix, validate, and move on. The buffer was what made that work. Today, that buffer is gone. AI didn't make your team slower. It changed the other side of the equation, compressing discovery-to-exploit from months to hours . And the sad truth for defenders is that a process built for breathing room can't survive without it. AI Turned Vulnerability Discovery Into a Volume Game In its May 2026 update, Anthropic reported that it and approximately 50 partners used Claude Mythos Preview to find more than 10,000 high- or critical-severity vulnerabilities in systemically important software in a single month. Earlier figures were just as stark. Pointed at Firefox, the gated Mythos model wrote 181 working exploits , against just 2 from t...

On June 9, Anthropic released Claude Fable 5 , the most capable model it has ever made, generally available. It also did something unusual: it shipped one model as two products, split not by capability but by a layer of safety classifiers. Fable 5 goes to the public. Its twin, Claude Mythos 5, the same underlying model with the cyber safeguards lifted, stays locked to a vetted group of cyber defenders and critical infrastructure operators. Anthropic calls Mythos 5 the strongest cybersecurity model in the world. The practical difference is this: Fable 5 routes flagged cyber, biology, chemistry, and distillation requests to the weaker Claude Opus 4.8, while Mythos 5 keeps the cyber capabilities available for vetted users. Both models cost $10 per million input tokens and $50 per million output tokens, less than half the price of the earlier Mythos Preview, and Fable 5 is available through the Claude API now. It is included on Pro, Max, Team, and seat-based Enterprise plans at no...

Meta on Tuesday announced that it will use information shared by other businesses to personalize users' feed and responses from its artificial intelligence (AI) chatbot, expanding its scope beyond targeted ads. "Businesses often share information about people's activity on their sites with us to make ads more relevant," Meta said in a statement. "We already use this data - like games you play or purchases you make on other websites - to make the ads you see more relevant. In the future, we'll use this information to personalize other parts of your experience, including the content you see in your Feed and AI responses." The social media giant emphasized that it's not collecting any new data as part of the update, adding users are in the driver's seat and that they get to decide how this information is used for personalization. To that end, Meta is streaming its controls by expanding the "Activity from other businesses" setting...

University of Toronto researchers have built and tested a proof-of-concept AI-driven computer worm that uses a locally hosted open-weight large language model to reason its way through a network, generate tailored attack strategies for each target it encounters, and replicate itself, all without human intervention and without touching a commercial AI service. The preprint, posted to arXiv on June 2 and currently under peer review, shows why single-CVE patching breaks down when malware can inspect exposed services, read fresh advisories, and generate a new attack path at runtime. In 15 isolated runs on a deliberately vulnerable 33-host network, the worm identified an average of 31.3 vulnerabilities and gained elevated access on 23.1 hosts, roughly three-quarters of the hosts it actively targeted. It then replicated autonomously to 20.4 of those hosts, or 62% of the full network, over seven days, with no prior knowledge of the network topology and no human input. Traditional worm...

Organizations have more visibility than ever. Growing tech stacks provide greater coverage, and network security teams are increasingly adopting AI and automation to help with routine tasks and reduce manual effort. But the same challenges persist. Outages still last hours, causing significant financial losses, operational disruption, and reputational impact. Threat response and mean time to remediate (MTTR) remain slow. Misconfigurations and human error still create major incidents. And, despite the promises of AI, teams remain overwhelmed and burnt out. Detection isn't the issue. Neither is tooling. Today, the real problem is execution - that is, the work that happens between tools. The hidden operational layer most organizations overlook Every time an alert fires, network security teams must: Gather context across systems Validate ownership and severity Route tickets to the appropriate people Request approvals Implement changes manually Log evidence...

The Miasma supply chain campaign has sparked a fresh attack wave called Hades , this time involving 37 malicious wheel artifacts across 19 packages in the Python Package Index (PyPI) registry, as the Mini Shai-Hulud-style attacks continue to be refined and splintered to target specific ecosystems. "The compromised releases shipped a *-setup.pth file that attempts to execute automatically during Python startup, download the Bun JavaScript runtime, and run an obfuscated JavaScript payload named _index.js," Socket said in a new analysis. The list of identified packages is below - bramin 0.0.2, 0.0.3, 0.0.4 cmd2func 0.2.2, 0.2.3 coolbox 0.4.1, 0.4.2 dynamo-release 1.5.4 executor-engine 0.3.4, 0.3.5 executor-http 0.1.3, 0.1.4 funcdesc 0.2.2, 0.2.3 magique 0.6.8, 0.6.9 magique-ai 0.4.4, 0.4.5 mrbios 0.1.1, 0.1.2 napari-ufish 0.0.2, 0.0.3 nucbox 0.1.2, 0.1.3 okite 0.0.7, 0.0.8 pantheon-agents 0.6.1, 0.6.2 pantheon-toolsets 0.5....

The U.S. Cybersecurity and Infrastructure Security Agency (CISA) on Monday added a high-severity flaw impacting BerriAI LiteLLM to its Known Exploited Vulnerabilities ( KEV ) catalog, citing evidence of active exploitation. The vulnerability, tracked as CVE-2026-42271 (CVSS score: 8.7), is a command injection vulnerability that could allow any authenticated user to run arbitrary commands on the host. It affects the following version of the LiteLLM Python package - >= 1.74.2 < 1.83.7 "Two endpoints used to preview an MCP server before saving it - POST /mcp-rest/test/connection and POST /mcp-rest/test/tools/list - accepted a full server configuration in the request body, including the command, args, and env fields used by the stdio transport," according to a description of the flaw shared by BerriAI. "When called with a stdio configuration, the endpoints attempted to connect, which spawned the supplied command as a subprocess on the proxy host w...

Phishing has always been a numbers game. AI has turned it into a volume machine. Attackers can now create convincing emails, fake login pages, and tailored lures in minutes. Every polished message adds another case for Tier 1 to review, another link to inspect, and another alert that cannot be dismissed at a glance. As the queue grows, a credential theft attempt or malware delivery can easily get buried among routine checks. SOC leaders need to help their teams cut through the noise faster and catch the alerts that could turn into a serious incident. Where Tier 1 Teams Lose Time on AI Phishing AI helps attackers launch more convincing campaigns, vary the message, and rotate infrastructure faster. For Tier 1 teams, that means fewer alerts can be ruled out quickly. AI-driven change What Tier 1 has to deal with SOC impact More lure variations Similar campaigns no longer look identical. More alert...

Monday again. The weekend was meant to be quiet. It wasn't. Last week had poisoned packages, a broken AI helper, and a worm tearing through repos. The ugly part: basic tricks still worked. A chatbot got fooled. A bot token got leaked inside the malware. The same old mistakes showed up again. And while everyone chased the loud stuff, quieter attackers sat in inboxes for months, reading mail and stealing it bit by bit. Lots to cover. Grab coffee. Read up. ⚡ Threat of the Week Miasma Worm Hits 73 Microsoft GitHub Repositories in Supply Chain Attack - Microsoft's GitHub repositories became the latest to fall victim to the ongoing Miasma self-replicating supply chain attack campaign. The incident impacted 73 Microsoft repositories across four of its GitHub organizations, including Azure, Azure-Samples, Microsoft, and MicrosoftDocs. The development prompted GitHub to disable access to those repositories. Miasma is assessed to be a variant of the Mini Shai-Hulud worm that T...

Mythos is real. I know a big chunk of the industry thinks it's a marketing stunt, and I get why. I get it. But I've seen the findings, and they're bad. These aren't "whoops, this line right here is wrong, and that's RCE." They're novel combinations of a few dozen issues out of thousands of things every SAST scanner already finds, chained together into something much worse. It's real creativity, like Move 37. That's not a better scanner. That's a different category of threat. In some ways, it doesn't even matter. Even if this specific model were a hoax, the capability is coming regardless. Some days, I wish it were a hoax. We'd have more time. But you can believe me or not. The rest of this post is about what we do about it either way, and I'm getting started now. Washington has been tracking this for a while, but you can't regulate something most of the industry thinks is made up. Now that every boardroom is in preparat...

OpenAI has begun rolling out a new Lockdown Mode to ChatGPT for eligible personal accounts to reduce the risk of data exfiltration arising from prompt injection attacks . The feature is primarily designed for people and organizations that handle sensitive data and require stricter protection guarantees. Lockdown Mode is available to logged-in users across Free, Go, Plus, and Pro, and self-serve ChatGPT Business plans. "Lockdown Mode is an optional advanced security setting that limits many tools and capabilities in OpenAI products that can connect to the web or external services," OpenAI said . "It is designed to reduce the risk of data exfiltration from prompt injection attacks by limiting outbound network requests, at the expense of disabling or limiting some useful features." The safeguards are aimed at hardening the attack surface against prompt injections, which continues to be a "frontier" problem impacting all large language models (LLMs...