AI This Week

Microsoft has introduced Project Ire, an artificial intelligence prototype aimed at reversing malware — a task typically handled by skilled human security researchers. The system identifies malicious files by analyzing software at multiple levels, from binary code to control flow and behavioral patterns. In early tests, Project Ire correctly flagged 90% of malicious Windows driver files and maintained a low false positive rate of 2%, indicating its potential to streamline threat detection. Unlike traditional antivirus methods, this system investigates deeper, even identifying hard-to-detect malware linked to elite hacking groups. While still a prototype, Project Ire has shown promise in complementing human researchers. Microsoft plans to integrate it into its Defender platform.

According to a study by EY Ireland, venture capital investment in generative AI has reached $49.2 billion in the first half of 2025, surpassing the total for all of 2024. The average deal size has tripled to $1.5 billion as investors focus on later-stage companies. The US dominates the market, accounting for 97% of global deal value. Key growth areas include agentic AI and specialized real-world applications. The study also notes a shift towards companies building software on third-party foundation models.

China has accused Nvidia of embedding a backdoor into its H20 chip, stirring tensions between Beijing and the U.S. The Cyberspace Administration of China claims that the chip contains location tracking and remote shutdown capabilities, raising significant national security concerns. These allegations surface just as Nvidia resumes H20 exports to China, following a temporary lift of U.S. restrictions. However, specifics on the backdoor claims remain vague. Nvidia has yet to comment. Meanwhile, China is actively boosting its domestic AI chip production, reducing reliance on foreign manufacturers. U.S. lawmakers are divided over renewed H20 sales, fearing risks to national security. This high-stakes technology conflict highlights the growing scrutiny over export controls and strategic chip supply chains.

Major technology companies are seeing strong returns on their massive artificial intelligence investments. Despite concerns over ballooning capital expenditures, firms like Microsoft, Meta, Amazon, and Alphabet are experiencing increased demand for AI-powered services across cloud computing, digital advertising, and internet search. This has led to revenue growth and positive investor sentiment. AI is becoming a primary growth engine for Big Tech, with companies ramping up spending to meet soaring demand.

Inbenta has announced significant achievements for the first half of 2025, reporting a 50% year-over-year rise in new customer acquisitions and expanded traction in European markets. The company saw a 21% increase in revenue from existing clients through enhanced platform usage, professional services, and added offerings. Use of its AI-powered solutions surged, with enterprise integrations scaling by 54%, showcasing the platform's adaptability for demanding environments. Additionally, Inbenta improved profitability while advancing innovation and preparing a major platform update slated for later this year. CEO Melissa Solis highlighted the team's dedication and the platform's growing impact on leading enterprises. With its AI applications delivering measurable results, Inbenta positions itself for further growth.

A groundbreaking AI model, Hierarchical Reasoning Model (HRM), developed by Sapient Intelligence, proves smaller can be smarter. With only 27 million parameters, HRM tackles reasoning puzzles that stump massive models like ChatGPT. Unlike traditional Chain-of-Thought methods, which falter with missteps, HRM mimics the brain with two modules: a strategic "planner" and a rapid "worker." This novel approach enables deep reasoning in a single forward pass. HRM has delivered stunning results. On the ARC-AGI benchmark, it outperformed larger models, and in extreme Sudoku and maze challenges, it excelled where others failed entirely. Its efficiency promises cheaper deployment, faster training, and innovative possibilities for local AI systems.

Replica Studios’ cutting-edge demonstration reveals how generative AI is transforming video games. Set inside a simulation inspired by "The Matrix," non-playable characters (NPCs) powered by AI grappled with existential dilemmas after discovering their world wasn’t real. Their unscripted, human-like responses — created using tools like ChatGPT — stunned onlookers, hinting at both the immersive potential and ethical concerns of AI integration.The technology’s ability to enhance gameplay with realistic dialogue raises questions about the future of the gaming industry. While it promises richer worlds, it also foreshadows potential upheaval for voice actors, writers, and creatives, whose roles may be supplanted by AI. This thought-provoking advancement leaves the industry at a crossroads.

NICE has announced its acquisition of Cognigy, a leading enterprise provider of conversational and agentic AI. Known for innovation in the AI industry, Cognigy supports major brands like Adidas, Nestlé, and Toyota with advanced virtual agent solutions. This strategic move bolsters NICE’s portfolio, pairing Cognigy’s cutting-edge AI offerings with NICE's CXone Mpower Orchestrator. The acquisition positions NICE to drive advancements in AI-first customer experiences while strengthening its footing in global markets. Industry experts believe the deal could trigger significant consolidation within the AI customer experience space. Cognigy’s extensive customer base and reputation for high satisfaction amplify the deal’s significance, with implications for end-to-end CX workflows.

The White House has released a new AI Action Plan, which aims to bolster US dominance in artificial intelligence. The plan, unveiled at an event co-hosted by tech investors, aims to cut environmental regulations to speed up AI supercomputer construction and promote U.S.-made AI technologies globally. It includes executive orders to fast-track AI projects, expand tech exports, and address perceived liberal bias in AI systems.

YouTube Shorts is launching several generative AI features for creators. Among the highlights is an image-to-video tool that transforms static photos into dynamic six-second videos. Users can animate landscape photos, group shots, or even everyday images using AI-powered suggestions. For example, the tool can breathe life into objects like pedestrian signals, creating engaging visuals. Additionally, creators can explore new AI effects to turn doodles into art or selfies into imaginative scenarios, such as swimming underwater. Powered by Google's Veo 2 video generation model, these tools come with clear labels and SynthID watermarks to mark AI-generated content. The features are currently rolling out to select regions.