AI and Machine Learning

Swiss privacy company Proton has launched Lumo 2.0, a major upgrade to its zero-access encrypted AI assistant, as concerns mount over the security risks posed by mainstream AI tools in the enterprise. Unlike most AI assistants, which log user conversations and may use them to train future models, Lumo is built so that not even Proton can read what users type into it. The company argues this addresses a growing blind spot in enterprise security:...

Check Point Software has announced it is embedding OpenAI's frontier cyber capabilities directly into its customer-facing security products, becoming one of a select group of vendors accepted into OpenAI's Daybreak Cyber Partner Programme. The move represents a significant escalation in the deployment of advanced AI in enterprise security, not as a back-end research tool but as a core component of the defences that more than 100,000 organisations rely on every day. What the Daybreak Programme...

Security teams are facing a new and accelerating threat: attackers armed with sophisticated AI models that can chain together multiple application vulnerabilities into fully formed exploits within minutes. That is the stark reality underpinning a raft of new capabilities announced today by Black Duck for its Polaris Platform. The company is positioning the enhancements around what it calls 'Mythos readiness,' a reference to advanced AI models now being used by threat actors. The urgency is not...

Check Point Software has been approved as a member of OpenAI's Trusted Access for Cyber (TAC) programme and accepted into Daybreak, OpenAI's cybersecurity initiative for vetted security organisations, the company announced today. The move gives the security vendor access to GPT-5.5 with Trusted Access for Cyber for use in defensive security operations, supporting security teams with threat analysis, incident investigation and real-time detection engineering. The announcement comes as the threat landscape is increasingly shaped by...

A growing tension sits at the heart of enterprise AI deployments: organisations want agents to act autonomously, yet handing over passwords and API keys to automated systems represents a significant and largely unresolved security risk. Proton is now attempting to close that gap with the launch of Proton Pass for AI Agents, a capability that wraps credential sharing in a structured token framework built around the principle of least-privilege access. Announced today, the feature introduces...

Check Point has launched an agentic orchestration platform that can end the policy drift, stalled Zero Trust projects, and manual configuration backlogs that have plagued enterprise security teams for decades. The company's Agentic Network Security Orchestration Platform is built around autonomous AI agents that translate business intent directly into firewall policy, tighten configurations in real time, and respond to threats across multi-vendor environments without waiting for a human to raise a ticket. It's an ambitious...

A 2025 study in Communications of the ACM tested 1,276 participants’ ability to spot AI-generated images, audio, and video. Average accuracy was 51.2% – literally no better than a coin flip. Participants scored worse on synthetic content (38%) than on real content (64%).  The problem is that most people default to thinking what they see is real. As synthetic media becomes more common across business, media, and online communication, the need for tools that can...

In early April, Anthropic announced the release of Claude Mythos Preview, its most advanced AI model to date. Though trained as a general-purpose large language model, Mythos quickly distinguished itself through a notable capacity for complex, multi-step cybersecurity work – autonomously parsing massive codebases, detecting critical-severity software defects and identifying vulnerabilities that have eluded human researchers for decades. Anthropic describes its cybersecurity capabilities as “substantially beyond those of any model we have previously trained,” including...

Black Duck has announced the general availability of Black Duck Signal™, an agentic AI application security solution designed from the ground up to address the security challenges created by AI-native software development. The launch comes as AI coding assistants move from novelty to norm across enterprise software teams. Industry analysts predict that 90% of enterprise developers will be using AI coding tools by 2028, a shift that is fundamentally changing the volume, velocity, and nature...

KnowBe4 has expanded its Artificial Intelligence Defense Agents (AIDA) suite with the launch of a new AI-powered assessment tool designed to help organisations measure human cyber risk more accurately. The company has introduced the Custom SAPA (Security Awareness Proficiency Assessment) AI Agent, which generates tailored security awareness assessments based on an organisation’s own technology environment, policies and workflows. The move aims to address the limitations of traditional standardised testing by providing security leaders with data...