A newly identified attack category that injects malicious instructions into AI agents via compromised tool output is creating a significant security surface for organisations.
Breaking News
Latest Posts
International Guests Put FREELANDER to the Test in Wuhu
The FREELANDER division of global automaker Chery concluded an intensive dynamic vehicle evaluation program in…
FREELANDER Showcases AI and Digital Technology to International Guests
Automotive innovator FREELANDER hosted an extensive technological showcase in Wuhu, China, presenting an international delegation…
Eskimoz Introduces an Integrated Approach to SEO, AI Search Visibility and Digital Acquisition
European digital acquisition agency Eskimoz unveiled a comprehensive search optimization methodology, integrating conventional technical search…
OpenAI and Broadcom’s Jalapeño Chip Physically Delivered to Sam Altman — Targets 50% Inference Cost Cut
The first custom AI inference chip designed by OpenAI and manufactured by Broadcom has been delivered as a physical unit targeting dramatically lower costs for running ChatGPT.
Nobel Laureate John Jumper Joins Anthropic From Google DeepMind — Latest in AI’s Historic Talent Exodus
AlphaFold lead and Nobel Prize winner John Jumper is the second major DeepMind scientist in a week to leave for a competitor as Anthropic continues its recruitment drive.
OpenAI Codex Reaches 5 Million Weekly Active Users — Samsung Adoption Boosts South Korean Growth 800%
Codex’s rapid enterprise growth is illustrated by Samsung’s adoption which drove weekly active users in South Korea up by nearly 800% since February 2026.
China’s Qwen AI Model Allegedly Benefited From 28.8 Million Distillation Attacks on Anthropic’s Claude
Anthropic’s Senate letter claims the massive extraction campaign was specifically designed to help Alibaba’s Qwen model approach capabilities of the restricted Mythos Preview.
Gemini 3.5 Pro Quietly Delayed From June to July — Google Misses Its Own Announced Launch Window
Google’s next major Gemini model has been pushed to July despite CEO promises of a June release adding to the list of frontier models that slipped past their announced dates.
GPT-5.6 Introduces Sol Terra and Luna — Three Models Targeting Frontier Reasoning Everyday Use and Speed
OpenAI’s new model family covers the full spectrum from frontier reasoning with Sol to affordable speed with Luna giving enterprises and developers more options for deployment.
EU AI Act High-Risk System Enforcement Deadline Approaches — Companies Race to Achieve Compliance
With the EU’s landmark AI regulation enforcement window closing organisations across Europe are rushing to audit their AI systems and implement the required safeguards.
Microsoft 2026 AI in Education Report Reveals Schools Moving From Experimentation to Real AI Implementation
Microsoft’s third annual AI education report finds widespread adoption and growing teacher confidence in AI tools accompanied by concerns about governance and equity.
Samsung Adopts OpenAI Codex as Company-Wide AI Platform Reversing 2023 Ban After Internal Code Leak
Samsung’s rollout of OpenAI tools across R&D manufacturing and marketing reverses a company-wide ChatGPT ban imposed after proprietary code was accidentally shared with the AI.