Artificial Intelligence | Breaking Cybersecurity News | The Hacker News

AI agents have become incredibly useful across many industries by their ability to reach inboxes, documents, and financial information, then search, summarize, or act on what they find. But that access creates a data-in-use problem: conventional encryption protects information in storage and transit, but an agent generally needs it decrypted in memory while processing it, where sensitive material can be exposed to application code, logs and debugging systems, infrastructure operators, or a compromised host. Conan Yu's work offers one practical response to that problem. He is co-founder of Rena Labs , which develops infrastructure for confidential AI training and inference using trusted execution environments (TEEs), hardware-isolated environments designed to limit the surrounding host's access while code and data are processed. Over roughly two years, Yu has worked across hardware and cloud TEE deployments, privacy-preserving financial-data analysis, and private AI inference. He a...

Most security teams know that dwell time matters. The harder question is what to do about it. Dwell time is the period between an attacker gaining access and the security team containing the threat. During that window, a threat actor has time to learn the environment, steal credentials, move between systems, and reach sensitive data. For years, security teams have tried to reduce this window by adding more detection tools. The problem is that more alerts do not necessarily mean faster detection.  A recent industry incident response report puts the global median dwell time at 14 days, up from 11 the year before, quietly reversing a run of steady improvement that had held for close to a decade. The better way to think about it is as an operational problem. Two numbers matter most. Mean time to detect (MTTD) tells you how quickly the team recognizes a real threat, while mean time to respond (MTTR) tells you how quickly the team investigates and contains it. An AI native SIEM...

As organizations deploy AI coding agents across large monorepos and microservices environments, a fundamental problem emerges: the model may be capable of making the change, yet still lack the organizational context required to make the right change safely. A developer can ask an AI coding agent to deprecate an API field, update an authentication flow, or modify a service interface. The agent can inspect the code available on the developer's machine and search for references. What it may not know is that the field is consumed by four other services across separate repositories, that one of those services belongs to another team, or that the same field eventually carries sensitive data into a third party integration. This is not simply a context window problem. It is a code context problem: providing AI agents with accurate, current, organization wide evidence about how software actually behaves. One emerging approach is to generate that evidence directly from source code us...

AI coding agents are already inside engineering organizations. The problem security teams need to solve is not only that AI-generated code might be vulnerable. You already have ways to catch that: code review, CI, SAST, dependency scanning, and production monitoring. The real problem is that tools such as Claude Code, OpenAI Codex, Claude Cowork, and GitHub Copilot are becoming extensible agent runtimes. Skills, plugins, hooks, repository instructions, and MCP servers can influence what the agent reads, which tools it selects, what commands it runs, and where enterprise data is sent. Most AI governance programs stop at approving the application. Very few can tell you everything that has been installed inside it. That is the supply-chain gap. What changed: Third-party components are no longer participating only at build or deploy. They are participating in the agent's decision loop. From coding assistant to agent runtime The first generation of coding assistants mainly...

Many AI systems depend on input signals that teams cannot fully inspect or explain. These opaque sources reduce visibility into the data paths that influence model behavior. Engineers lose provenance records, limiting the diagnosis of abnormal outputs. This complicates the work of security teams that need clear records of what influenced a model at any point in time. Verifiable search data offers a stable alternative. It gives teams an input they can examine, store, and reproduce in controlled conditions. Engineers can compare model behavior against information that was publicly accessible at the time a result was produced, rather than depend on hidden internal signals. This article outlines why verifiable search data gives AI and security teams the clarity required to maintain operational control. Why Traceability Matters in AI Systems Traceability lets teams follow an input from its origin through each processing step. When every stage can be inspected, engineers can review...

When it comes to email security software, the cybersecurity industry has become very good at one thing: scoring the message that's already in the inbox. We've gone from blocklists to signature matching to behavioral ML, and each generation of email security innovation was a genuine improvement over the last.  But the losses keep climbing anyway. The median time it takes someone to click a phishing link is just 21 seconds after opening the email, and another 28 seconds to hand over credentials or payment data. That's under a minute start to finish. That number hasn't moved much in years despite everything we've thrown at the problem. It's time for a different approach. But first, we need to understand modern challenges. We're filtering messages, but attackers are running campaigns

Somewhere right now, malware running on a compromised machine is checking in with an AI model, asking it for a new version of itself. Google's Threat Intelligence Group caught this happening in late 2025. The malware, nicknamed PROMPTFLUX, does this every hour it runs, and each version comes back looking different from the last. By the time a security tool learns to recognize it, it has already changed shape again. This isn't a rare glitch or a lab experiment. It's a preview of how a growing share of attacks work today and why legacy SIEM platforms built to detect known patterns are starting to fall behind. Legacy security tools were built to detect, not to adapt A SIEM is a system that collects logs from every part of a company's network and looks for signs of an attack. For years, it worked like a security guard with a very long memory. The guard learns what a break-in looks like - a certain kind of file, a pattern of behavior, a code signature, and watches fo...

For a long time, the cybersecurity industry has framed social engineering as a psychological problem. We treat it as a battle of wits between a charismatic con artist and an unsuspecting employee. The prevailing wisdom says that if we just train our people to better identify scams or tear down malicious infrastructure slightly faster, we can stay ahead of scammers. But looking at the threat landscape from a purely threat-intelligence perspective tells us something entirely different: Modern social engineering is more than a psychological game. It's a highly optimized, industrialized deception economy. Attackers run campaigns like hyper-efficient businesses. They have Customer Acquisition Costs, operational budgets, and strict Return on Investment (ROI) targets. This is the part the industry doesn't like to sit with: If we want to truly break the social engineering attack chain , we have to stop focusing exclusively on building higher walls or executing reactive takedowns...

For years, our industry has treated "blocking" as the gold standard. If the email didn't land, if the malware didn't execute, if the alert fired in the SIEM, we called it a win. That mindset made sense in a world where most attacks came through a handful of familiar doors. But AI has changed the game. We're not dealing with hobbyists sending out clumsy phishing attacks anymore. Modern adversaries are running multi‑channel, AI‑assisted businesses at machine speed. And if all you're doing is blocking at the edge, you're not really defending. You're just delaying. Generative AI has made it trivial to spin up highly personalized, multi‑step social engineering campaigns that operate simultaneously across email, collaboration apps, mobile, social media, and paid media. The result is a social engineering attack chain : a sequence of stages designed to manufacture trust, erode judgment, and bypass brittle controls. You don't beat that by tuning another filter. You have to disrupt the at...

AI is changing the security landscape. More and more threat groups incorporate LLMs into their reconnaissance and exploitation workflows. The notion that some vulnerabilities are too complex to implement is now obsolete. Using LLMs, hackers can automatically find and exploit complex vulnerabilities. We have all heard of Claude Mythos and its ability to identify vulnerabilities in large codebases and exploit them automatically. But LLMs can do more than find vulnerabilities in code. ShinyHunters has scanned thousands of Salesforce Sites. They used a modified version of "AuraInspector". They possibly used an LLM to code their framework, mods, reconnaissance tools, and other aspects of their workflow. But the next step is to use AI to supercharge the attack process itself. We at Reco decided to explore what it would look like. Reco's security research team built an AI-powered agent capable of performing end-to-end security assessments of Salesforce Experience Cloud sit...

AI-driven vulnerability discovery is no longer a research project. Claude Mythos proved that. In a single sweep, it uncovered thousands of vulnerabilities in software we use every day, generated working exploits, and exposed bugs that had survived decades of human review. Other AI models are rapidly catching up, and we've entered into an entirely new operating environment for cybersecurity. The industry is treating this as a turning point, and it is. But not for the reason most people might think. The Real Problem Was Never Finding Vulnerabilities Most of the conversation around AI security focuses on discovery: AI can now identify vulnerabilities faster than human teams ever could. That is certainly true, but it also misses the larger operational reality organizations have been struggling with for years. Security teams were already overwhelmed long before AI entered the picture. Vulnerability scanners, fuzzers, and static analysis tools have consistently generated more...

Phishing emails have reached a point where they can fool both people and the tools designed to stop them. For anyone working through a packed inbox, it's easy to trust what looks familiar and click without a second thought. What's worrying is that phishing is rarely the end goal. It's usually the entry point for something much bigger: a ransomware attack. Once attackers gain access, they don't act immediately. They move through systems, map connections, and prepare the environment. By the time ransomware is deployed, it's the final step — not the first. To stay ahead, you need protection at two critical points. An advanced email security solution that catches even the most stealthy phishing attempts, and a strong BCDR strategy that lets you restore data quickly and avoid paying a ransom if something slips through. Why phishing remains so effective Phishing works because it plays on human behavior. Email may seem like a simple communication tool, but it functions as a decision-mak...

Most of the commentary on Anthropic's Claude Mythos Preview has gone in one of two directions: one camp treats it as the civilizational inflection point, the other as marketing dressed up as a research result. Neither read is particularly useful for a security leader who still has a program to run on Monday. The AISLE team's technical response to the Mythos announcement made a fair point worth sitting with: much of what was demonstrated is recoverable on smaller, open-weight models, particularly on the discovery side. Early testing results of OpenAI's GPT 5.5 show CTF performance close to or slightly superior to Mythos; the exclusivity framing is arguable, but the accelerated model improvement in offensive security is undisputable. The UK AI Security Institute found that Mythos can autonomously execute a complete corporate network takeover, succeeding in 30% of its attempts on a complex attack range — a task AISI estimates would require roughly 20 hours for a human e...

For security leaders, the inbox remains the front door for attackers. Here's why the smartest teams are adding adaptive, AI-driven protection to their cloud email security, not replacing them. Email is still the number-one attack vector for enterprises, and it is not even close. The FBI's Internet Crime Complaint Center reported that business email compromise alone generated $3 billion in losses in 2024 , with AI-enabled attacks accelerating the trend ( FBI IC3 Report ). The attacks that succeed today don't carry obvious malicious payloads. They rely on trust, tone, and timing; a spoofed vendor sending a "routine" invoice update, or a convincing impersonation of a CEO with an urgent request. No malware. No suspicious links. Just words, carefully chosen. Microsoft 365 is the backbone of productivity for most organizations, and Microsoft Defender and Exchange Online Protection do solid work catching known spam, malware, and co...

We recently worked with an organization that had invested heavily in advanced security tooling, including AI-driven detection and monitoring capabilities. From a technical perspective, the environment appeared mature: alerts were firing, dashboards were populated, and risks were clearly identified.  Yet progress had stalled.  The security team and IT disagreed on ownership. Business leadership perceived cyber risk as "under control," while the security team felt increasingly exposed and unheard. AI surfaced the signals, but no one could agree on what to do with them.  The turning point did not come from additional tooling or deeper analysis. It came from reframing the conversation.  By aligning stakeholders around clear business impact, contextualizing the findings against industry peers, and translating technical gaps into credible, board-level risk narratives that reinforced the internal security team's concerns rather than questioning their judgment, decisions were finally ma...

While working on the Transparent Tribe's vibeware research, we have encountered two distinct camps, the optimists and the skeptics. What makes the current dialogue unique is that both sides can be right at the same time. There is, however, a clear operational reason why we encounter "AI attacks" primarily on professional social media feeds rather than within our own telemetry logs. In this article, we analyze the factors explaining why Skynet is not here yet, and how, much like a shark, AI does not need to be innovative to be dangerous. LLM Architecture Bias LLMs are mathematically optimized to predict the most likely outcome, while hacking is the art of identifying the statistical anomaly. LLMs are designed to predict the most statistically probable next token. They are excellent at the average, but poor at the exceptional. A hacker, by contrast, is a practitioner of statistical anomaly, actively seeking the low-pro...

The conversation around AI in the SOC has mostly centered on efficiency: closing alerts faster, reducing queue backlog, and automating repetitive work that burns out L1 analysts. That framing is directionally right, and it matters because analyst fatigue is real. For teams dealing with high alert volume, analysts are often asked to make good decisions under a fragmented context and time pressure. But that framing is still incomplete. The bigger shift is not just workflow automation or orchestration of predefined playbooks. It is AI's ability to perform contextual, hypothesis-driven investigation across multiple telemetry sources, work that has traditionally depended on experienced L2 or L3 analysts and limited human time. When that capability can be applied consistently across every alert, it changes the operating model, not just the speed of the existing one. Two recent investigations at Prophet Security make that real. In both cases, the attacks were not obvious from signature-bas...

AI hype is everywhere. Every security vendor claims their platform is "AI-powered." Dashboards promise automation. Generative AI is positioned as the solution to staffing shortages. For small and mid-sized organizations with lean IT and cybersecurity teams, these messages are understandably compelling. But this leads to a critical question: Can AI realistically strengthen your security program — and is it worth the effort? The Current Reality: Under-Resourced and Overwhelmed Small and midsized organizations face a difficult equation. Threat actors are becoming more sophisticated. Attack surfaces continue to expand. Compliance pressures are rising. Meanwhile, security teams are small — often just a few professionals wearing multiple hats. AI sounds like a relief. In theory, it can accelerate detection, reduce alert fatigue, automate triage, improve response times, and surface hidden threats buried in large volumes of data. But AI is not plug-and-play magic for defenders. For l...

In the digital world, the secure exchange of cryptographic keys is the foundation upon which all private communication is built. It's the initial, critical handshake that allows two parties, like a user's browser and a web server, to establish a shared secret and communicate securely over the untrusted expanse of the internet. As the quantum computing era approaches, the very mathematics underpinning our traditional key exchange mechanisms are facing an existential threat. This spurred the development of new, quantum-resistant algorithms. This blog post provides a deep dive into how modern key exchange works, from the trusted classical methods to the emerging post-quantum standards, and explores how Zscaler leverages hybrid key exchange to bridge the gap. The Key Components of Modern Key Exchange At a high level, a secure key exchange protocol must achieve the following: Confidentiality: The established key must be a secret shared only between the two communicating parties. An ea...