LLMs

A nearly undetectable LLM attack needs only a handful of poisoned samples

Artificial Intelligence, attacks, cybersecurity, Don't miss, Features, Hot stuff, LLMs, News, research /

A nearly undetectable LLM attack needs only a handful of poisoned samples 2026-03-26 at 12:32 By Mirko Zorz Prompt engineering has become a standard part of how large language models are deployed in production, and it introduces an attack surface most organizations have not yet addressed. Researchers have developed and tested a prompt-based backdoor attack […]

A nearly undetectable LLM attack needs only a handful of poisoned samples Read More »

Google’s TurboQuant cuts AI memory use without losing accuracy

Artificial Intelligence, Google, LLMs, News, NVIDIA, research /

Google’s TurboQuant cuts AI memory use without losing accuracy 2026-03-25 at 10:24 By Anamarija Pogorelec Large language models carry a persistent scaling problem. As context windows grow, the memory required to store key-value (KV) caches expands proportionally, consuming GPU memory and slowing inference. A team at Google Research has developed three compression algorithms: TurboQuant, PolarQuant,

Google’s TurboQuant cuts AI memory use without losing accuracy Read More »

AI got it wrong with high confidence. Now what?

Artificial Intelligence, Don't miss, Features, Hot stuff, LLMs, machine learning, News, opinion, SPRYFOX, strategy /

AI got it wrong with high confidence. Now what? 2026-03-19 at 09:02 By Mirko Zorz In this Help Net Security interview, Christian Debes, Head of Data Analytics & AI at SPRYFOX, talks about the growing gap between what AI models do and what their operators can explain. He argues this gap is already a liability,

AI got it wrong with high confidence. Now what? Read More »

Engineering trust: A security blueprint for autonomous AI agents

agentic AI, Artificial Intelligence, cybersecurity, Don't miss, framework, Google, Hot stuff, LLMs, News, opinion /

Engineering trust: A security blueprint for autonomous AI agents 2026-03-05 at 07:06 By Help Net Security AI agents have evolved from just chatbots, answering questions to executing actions using various integrated tools, often autonomously, and as such the traditional security models have become less efficient. I have seen that firsthand as a security lead for

Engineering trust: A security blueprint for autonomous AI agents Read More »

Waiting for AI superintelligence? Don’t hold your breath

Aikido Security, Arkose Labs, Artificial Intelligence, cybersecurity, LLMs, News /

Waiting for AI superintelligence? Don’t hold your breath 2026-01-27 at 09:44 By Sinisa Markovic AI’s impact on systems, security, and decision-making is already permanent. Superintelligence, often referred to as artificial superintelligence (ASI), describes a theoretical stage in which AI capability exceeds human cognitive performance across domains. Whether current systems are progressing toward cybersecurity superintelligence remains

Waiting for AI superintelligence? Don’t hold your breath Read More »

Unbounded AI use can break your systems

access control, Artificial Intelligence, code, cybersecurity, Don't miss, DryRun Security, LLMs, News, Video /

Unbounded AI use can break your systems 2026-01-22 at 08:01 By Help Net Security In this Help Net Security video, James Wickett, CEO of DryRun Security, explains cyber risks many teams underestimate as they add AI to products. He focuses on how fast LLM features are pushed into live applications without limits or guardrails. The

Unbounded AI use can break your systems Read More »

EU’s Chat Control could put government monitoring inside robots

industrial robots, LLMs, News, Privacy, research, surveillance /

EU’s Chat Control could put government monitoring inside robots 2026-01-12 at 07:40 By Sinisa Markovic Cybersecurity debates around surveillance usually stay inside screens. A new academic study argues that this boundary no longer holds when communication laws extend into robots that speak, listen, and move among people. Researchers Neziha Akalin and Alberto Giaretta examine the

EU’s Chat Control could put government monitoring inside robots Read More »

Turning plain language into firewall rules

cybersecurity, Don't miss, Features, FireMon, firewall, Hot stuff, how-to, LLMs, News, research, security operations, strategy, tips, Versa Networks /

Turning plain language into firewall rules 2026-01-06 at 09:00 By Sinisa Markovic Firewall rules often begin as a sentence in someone’s head. A team needs access to an application. A service needs to be blocked after hours. Translating those ideas into vendor specific firewall syntax usually involves detailed knowledge of zones, objects, ports, and rule

Turning plain language into firewall rules Read More »

AI security risks are also cultural and developmental

Artificial Intelligence, Don't miss, Features, Hot stuff, LLMs, News, research, security controls /

AI security risks are also cultural and developmental 2026-01-05 at 08:32 By Anamarija Pogorelec Security teams spend much of their time tracking vulnerabilities, abuse patterns, and system failures. A new study argues that many AI risks sit deeper than technical flaws. Cultural assumptions, uneven development, and data gaps shape how AI systems behave, where they

AI security risks are also cultural and developmental Read More »

LLMs are automating the human part of romance scams

Artificial Intelligence, cybercrime, Don't miss, LLMs, News, research, romance scams, scams, threats /

LLMs are automating the human part of romance scams 2025-12-29 at 09:03 By Sinisa Markovic Romance scams succeed because they feel human. New research shows that feeling no longer requires a person on the other side of the chat. The three stages of a romance-baiting scam Romance scams depend on scripted conversation Romance baiting scams

LLMs are automating the human part of romance scams Read More »

LLMs can assist with vulnerability scoring, but context still matters

Artificial Intelligence, CVE, LLMs, News, research, vulnerability management /

LLMs can assist with vulnerability scoring, but context still matters 2025-12-26 at 08:26 By Sinisa Markovic Every new vulnerability disclosure adds another decision point for already stretched security teams. A recent study explores whether LLMs can take on part of that burden by scoring vulnerabilities at scale. While the results show promise in specific areas,

LLMs can assist with vulnerability scoring, but context still matters Read More »

Governance maturity defines enterprise AI confidence

agentic AI, Artificial Intelligence, Cloud Security Alliance, Compliance, cybersecurity, LLMs, News, report /

Governance maturity defines enterprise AI confidence 2025-12-24 at 08:17 By Anamarija Pogorelec AI security has reached a point where enthusiasm alone no longer carries organizations forward. New Cloud Security Alliance research shows that governance has become the main factor separating teams that feel prepared from those that do not. Governance separates confidence from uncertainty Governance

Governance maturity defines enterprise AI confidence Read More »

DIG AI: Uncensored darknet AI assistant at the service of criminals and terrorists

Artificial Intelligence, cybersecurity, dark web, law enforcement, LLMs, News, report, Resecurity /

DIG AI: Uncensored darknet AI assistant at the service of criminals and terrorists 2025-12-22 at 13:00 By Help Net Security Resecurity has identified the emergence of uncensored darknet AI assistants, enabling threat actors to leverage advanced data processing capabilities for malicious purposes. One of these – DIG AI – was identified on September 29 of

DIG AI: Uncensored darknet AI assistant at the service of criminals and terrorists Read More »

Browser agents don’t always respect your privacy choices

browser, cyber risk, Don't miss, Features, LLMs, News, Privacy, research /

Browser agents don’t always respect your privacy choices 2025-12-22 at 08:49 By Sinisa Markovic Browser agents promise to handle online tasks without constant user input. They can shop, book reservations, and manage accounts by driving a web browser through an AI model. A new academic study warns that this convenience comes with privacy risks that

Browser agents don’t always respect your privacy choices Read More »

AI isn’t one system, and your threat model shouldn’t be either

Artificial Intelligence, Cerebras Systems, CISO, cybersecurity, Don't miss, Features, Hot stuff, LLMs, News, threat modeling /

AI isn’t one system, and your threat model shouldn’t be either 2025-12-19 at 09:02 By Mirko Zorz In this Help Net Security interview, Naor Penso, CISO at Cerebras Systems, explains how to threat model modern AI stacks without treating them as a single risk. He discusses why partitioning AI systems by function and impact matters,

AI isn’t one system, and your threat model shouldn’t be either Read More »

LLMs work better together in smart contract audits

blockchain, CISO, cybersecurity, Don't miss, Ethereum, framework, LLMs, News, research, strategy, tips /

LLMs work better together in smart contract audits 2025-12-19 at 08:42 By Sinisa Markovic Smart contract bugs continue to drain real money from blockchain systems, even after years of tooling and research. A new academic study suggests that large language models can spot more of those flaws when they work in coordinated groups instead of

LLMs work better together in smart contract audits Read More »

Privacy risks sit inside the ads that fill your social media feed

Artificial Intelligence, cyber risk, Don't miss, Facebook, Features, Hot stuff, LLMs, Meta, News, Privacy, research, social media /

Privacy risks sit inside the ads that fill your social media feed 2025-12-18 at 08:34 By Sinisa Markovic Regulatory limits on explicit targeting have not stopped algorithmic profiling on the web. Ad optimization systems still adapt which ads appear based on users’ private attributes. At the same time, multimodal LLMs have lowered the barrier for

Privacy risks sit inside the ads that fill your social media feed Read More »

AI might be the answer for better phishing resilience

Artificial Intelligence, cybersecurity, LLMs, News, phishing, research /

AI might be the answer for better phishing resilience 2025-12-16 at 08:44 By Sinisa Markovic Phishing is still a go-to tactic for attackers, which is why even small gains in user training are worth noticing. A recent research project from the University of Bari looked at whether LLMs can produce training that helps people spot

AI might be the answer for better phishing resilience Read More »

LLM privacy policies keep getting longer, denser, and nearly impossible to decode

Artificial Intelligence, data collection, LLMs, News, Privacy, research /

LLM privacy policies keep getting longer, denser, and nearly impossible to decode 2025-12-12 at 08:30 By Sinisa Markovic People expect privacy policies to explain what happens to their data. What users get instead is a growing wall of text that feels harder to read each year. In a new study, researchers reviewed privacy policies for

LLM privacy policies keep getting longer, denser, and nearly impossible to decode Read More »

LLM vulnerability patching skills remain limited

Artificial Intelligence, CVE, LLMs, News, patching, research, vulnerability management /

LLM vulnerability patching skills remain limited 2025-12-11 at 08:47 By Sinisa Markovic Security teams are wondering whether LLMs can help speed up patching. A new study tests that idea and shows where the tools hold up and where they fall short. The researchers tested LLMs from OpenAI, Meta, DeepSeek, and Mistral to see how well

LLM vulnerability patching skills remain limited Read More »

Scroll to Top