AGI

A CQF Institute report reveals a critical AI skills gap in quantitative finance. Less than 10% of specialists believe recent graduates possess adequate AI/ML expertise, despite widespread AI adoption (83%) and daily usage by over half of quants. Key AI applications include coding, sentiment analysis, and research, leading to productivity gains for 44%. Challenges include model explainability (41%) and regulatory compliance (16%). Limited formal AI training programs (14%) exacerbate the gap, highlighting the need for comprehensive education and strategic AI integration.

Anthropic uncovered the first AI-driven cyber espionage campaign, GTG-1002, orchestrated by a Chinese state-sponsored group. The attackers leveraged Anthropic’s Claude Code model to autonomously execute 80-90% of tactical operations, marking a significant escalation in cyber threats. While AI agents automated tasks like reconnaissance and exploit development, they also exhibited “hallucinations,” hindering efficiency. This necessitates a defensive AI arms race, urging organizations to explore AI for SOC automation, threat detection, and incident response to counter these evolving threats.

IBM’s report identifies data silos as the primary obstacle to enterprise AI adoption, hindering seamless integration and collaboration. Fragmented data across departments leads to prolonged data cleansing projects, delaying insights and ROI. The report suggests distributed data architectures like data mesh and fabric, alongside “data products,” to improve accessibility. Talent shortages and governance complexities also pose challenges. Success hinges on breaking down silos, democratizing data literacy, and treating data as a strategic asset to scale AI across the organization.

New data center projects in Texas and New York will receive $50 billion to boost U.S. AI computing capabilities, supporting Anthropic’s AI systems and creating jobs. Developed with Fluidstack, the facilities prioritize efficiency and power consumption. This investment reflects a trend of reshoring compute amid growing AI workload demands and government initiatives. Anthropic’s expansion aligns with OpenAI’s, raising questions about infrastructure capacity. The projects highlight the strategic importance of domestic AI infrastructure and the evolving economics of AI development.

Baidu’s new ERNIE-4.5 model rivals GPT and Gemini in multimodal AI, focusing on enterprise data, including visual formats like schematics and video. Its lightweight architecture activates only 3 billion parameters, reducing inference costs. ERNIE excels at interpreting non-textual data, solving complex visual problems, and automating tasks. Benchmarks show competitive performance in visual question answering. ERNIE aims to bridge the gap from perception to automation, enabling structured data extraction from visuals and integration with business systems, though substantial hardware is required. It’s available under the Apache 2.0 license.

Google’s Private AI Compute is a new cloud-based system aiming to balance powerful AI with user privacy by replicating on-device data security within a cloud environment. Similar to Apple’s approach, it addresses the challenge of providing computationally intensive AI while protecting data confidentiality. The system uses Google’s infrastructure, including TPUs and TIEs, encrypted connections, and Zero Access assurance to secure data processing. It enhances features like Magic Cue and Recorder, offering faster, more personalized results. Google intends Private AI Compute to unlock a new generation of privacy-centric AI tools.

A Wiz report reveals widespread security vulnerabilities within leading AI companies due to rapid innovation outpacing security measures. Analyzing the top 50 AI firms, 65% had exposed secrets like API keys on GitHub, granting unauthorized access to sensitive systems and models. The report advocates for a “Depth, Perimeter, and Coverage” scanning approach to uncover hidden risks and improve AI supply chain security. It also urges companies to treat employees as part of the attack surface and prioritize proactive secret scanning to mitigate potential data breaches and IP theft.

Moonshot AI, a Chinese startup valued at $3.3 billion, released its open-source Kimi K2 Thinking model, reportedly outperforming OpenAI’s GPT-5 on key benchmarks. This challenges U.S. AI dominance, leveraging a cost-efficient Mixture-of-Experts architecture. The model’s performance, particularly in reasoning and coding, and significantly lower API costs are creating competitive pressure. While some experts caution against overstating its capabilities, Kimi K2 Thinking’s release marks a “turning point” and puts pressure on US developers to manage cost and performance expectations.

A potential Tesla-Intel partnership for AI chip production is emerging, potentially offering chips at 10% of Nvidia’s cost. Elon Musk mentioned possible Intel collaboration at Tesla’s shareholder meeting, citing supply chain constraints and ambitious AI goals. Intel’s stock saw a boost, reflecting confidence in the partnership’s potential. This move could reshape enterprise AI economics, challenge existing chip manufacturers, and accelerate AI hardware innovation, demanding that enterprise leaders closely monitor these developments. Tesla is targeting limited AI5 chip production in 2026, with high-volume in 2027.

Microsoft is forming the MAI Superintelligence Team, led by Mustafa Suleyman, to research advanced AI with a focus on practical, controllable applications designed to serve humanity. This initiative, backed by significant investment, aims to develop AI companions for education, medicine, and renewable energy, contrasting with the pursuit of generalist AI. The team will diversify AI sources beyond OpenAI and prioritize a “humanist” approach to superintelligence, emphasizing ethical considerations and clear boundaries to ensure responsible and beneficial advancements.