AI News selected for Professionals and Decision Makers
AI Security And Privacy Updates

Google Launches Gemini 3.5 Flash Cyber AI to Find and Fix Software Vulnerabilities

17:09 · July 21, 2026 · Hacker News AI Section

Google Launches Gemini 3.5 Flash Cyber AI to Find and Fix Software Vulnerabilities

Google's DeepMind on Tuesday announced the release of Gemini 3.5 Flash Cyber, a specialized artificial intelligence (AI) model built atop 3.5 Flash that's designed to discover, validate, and patch vulnerabilities quickly and efficiently. According to the tech giant, the model will be exclusively available to governments and trusted partners via CodeMender as part of a limited-access pilot

Summary

Google DeepMind has released Gemini 3.5 Flash Cyber, a specialized model built on the 3.5 Flash foundation and tuned for vulnerability discovery, validation, and patching. The lightweight design serves as a cost-efficient alternative to larger dedicated security models, allowing repeated high-speed calls that let an agent examine more code paths without proportional increases in expense.

Initial access is limited to a pilot through CodeMender, the AI-powered agent Google introduced in October 2025. The model is available only to governments and trusted partners, a deliberate restriction intended to give defenders an early advantage while limiting broader misuse of its dual-use capabilities. CodeMender runs the model inside controlled guardrails that enable defensive tasks such as forensic analysis while blocking offensive operations.

In internal evaluations the model outperformed both Gemini 3.5 Flash and 3.6 Flash on new vulnerability detection. On the V8 JavaScript engine it identified 55 unique confirmed issues under a fixed invocation budget, compared with 47 for 3.5 Flash and 36 for Anthropic Claude Opus 4.6, including ten issues missed by the other systems. It has also produced a fully reliable remote-code-execution exploit that bypassed standard mitigations including address-space layout randomization.

Google plans to extend the underlying CodeMender capabilities to enterprise customers through the Gemini Enterprise Agent Platform and to add red-teaming and end-to-end defense features over time.

Why it matters

This article is highly relevant for Dutch security professionals as it introduces a state-of-the-art AI tool for automated vulnerability discovery and patching. Given the strict EU regulatory landscape (like NIS2 and the Cyber Resilience Act), leveraging such AI capabilities will be critical for Dutch enterprises and government bodies to proactively secure software supply chains.

More in this beat
codemendergeminigooglemodel-security-controlsopus-4-6red-teamingthreat-and-vulnerability-updates
Three Recent Chrome Releases Fix 1,442 Flaws, More Than Prior 23 Updates Combined

14:51 · July 31, 2026

Three Recent Chrome Releases Fix 1,442 Flaws, More Than Prior 23 Updates Combined

This article highlights how AI and LLMs are fundamentally changing the cybersecurity landscape by accelerating vulnerability discovery and exploitation. Dutch security professionals must adapt their vulnerability management strategies to handle the increased volume of AI-driven threat disclosures in ubiquitous enterprise software.

Relevance 85 · Audience 95

Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

02:00 · July 21, 2026

Introducing Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber

Direct model update with actionable details on efficiency, cost, benchmarks, and integration for building AI agents. Product teams can evaluate token savings, latency, and coding/multimodal gains for production use. Includes limitations and safety considerations relevant to EU deployment.

Relevance 85 · Audience 90

Russian-Speaking Hacker Uses Google Gemini CLI to Control Botnet of Eight Dental Clinic PCs

11:07 · July 20, 2026

Russian-Speaking Hacker Uses Google Gemini CLI to Control Botnet of Eight Dental Clinic PCs

This article is highly relevant for security professionals as it demonstrates a real-world case of AI being weaponized to automate and manage cyberattacks. Understanding these AI-driven tactics is crucial for Dutch enterprises to update their threat models and develop countermeasures against highly adaptable, AI-assisted threat actors.

Relevance 85 · Audience 95

More details on Fable 5’s cyber safeguards and our jailbreak framework

02:00 · July 2, 2026

More details on Fable 5’s cyber safeguards and our jailbreak framework

Provides actionable, specific guidance on model-level cyber safeguards and a structured jailbreak evaluation rubric directly usable by product teams building or auditing AI systems, with clear discussion of dual-use risks and deployment trade-offs.

Relevance 85 · Audience 80

ThreatsDay: Gogs 10.0 RCE, n8n Workflow-to-RCE, $10M Reward, GLM-5.3 AI Exploit and More

19:23 · August 20, 2026

ThreatsDay: Gogs 10.0 RCE, n8n Workflow-to-RCE, $10M Reward, GLM-5.3 AI Exploit and More

The article provides crucial updates on privacy-enhancing technologies for AI that are vital for GDPR compliance in the Netherlands. It also alerts security professionals to emerging AI-driven threats, such as uncensored LLMs and AI models capable of autonomous vulnerability exploitation, which require immediate defensive consideration.

Relevance 85 · Audience 95

New Cryptographic Context Injection Attack Could Let Web Pages Steal Grok Chat Data

16:36 · August 20, 2026

New Cryptographic Context Injection Attack Could Let Web Pages Steal Grok Chat Data

This article highlights a critical data exfiltration vulnerability in LLMs via context injection, which is highly relevant for security professionals defending AI systems. Understanding this attack vector is essential for Dutch enterprises to ensure GDPR compliance and protect user privacy when deploying AI chatbots.

Relevance 85 · Audience 95

Reasoning Jury: Multi-Model Consensus for Evaluating Reasoning Traces

06:00 · August 15, 2026

Reasoning Jury: Multi-Model Consensus for Evaluating Reasoning Traces

Directly actionable for Dutch researchers and advanced practitioners building or fine-tuning reasoning LLMs; leverages open models to bypass closed-model guardrails, supporting EU transparency and ethical-AI requirements; high technical depth and reproducibility make it suitable for Primary research stream readers.

Relevance 82 · Audience 88

AI Exposes Enterprise Data via Prompt Injection

17:19 · August 13, 2026

AI Exposes Enterprise Data via Prompt Injection

Directly addresses AI-specific security risks and privacy threats with actionable recommendations on data governance and access controls, highly relevant for Dutch/EU security professionals managing AI deployments under GDPR.

Relevance 85 · Audience 90

AI Recommendation Poisoning: How "Ask AI" Buttons Silently Alter LLM Memory

13:30 · August 6, 2026

AI Recommendation Poisoning: How "Ask AI" Buttons Silently Alter LLM Memory

Directly addresses AI security risks from prompt injection and memory poisoning with actionable guidance for professionals. Applicable to Dutch/EU teams using commercial AI tools, aligning with GDPR and AI Act compliance needs. Provides concrete detection patterns and policy recommendations.

Relevance 85 · Audience 90

When Vibe Hacking Turns AI into the Junior Hacker Every Adversary Always Wanted

13:30 · August 4, 2026

When Vibe Hacking Turns AI into the Junior Hacker Every Adversary Always Wanted

This article is highly relevant for security professionals as it highlights the evolving AI-driven threat landscape where the technical barrier to entry for attackers is significantly lowered. It provides actionable strategic advice on shifting from point-in-time security assessments to continuous threat exposure management, which is crucial for Dutch enterprises defending against AI-assisted cyberattacks.

Relevance 85 · Audience 90