The startup Humans&, founded by alumni from major tech companies, has raised $480 million to develop a new AI model focused on enhancing collaboration and coordination among users.

Unlike traditional chatbots, this model aims to address the complexities of teamwork and decision-making. The founders emphasize the importance of social intelligence in AI, moving beyond mere information retrieval.

While the product is still in development, it is intended to improve communication and collaboration in both enterprise and consumer contexts, positioning itself as a central player in the evolving AI landscape.

👉 Read the full article on TechCrunch

Large language models (LLMs) increasingly generate code with minimal human oversight, raising critical concerns about backdoor injection and malicious behavior.

This research introduces the Cross-Trace Verification Protocol (CTVP), a novel AI control framework that detects anomalous behavior indicative of backdoors by analyzing consistency patterns in the model's predictions of execution traces across semantically equivalent program transformations.

The approach introduces the Adversarial Robustness Quotient (ARQ), which quantifies the computational cost of verification and shows exponential growth with orbit size. Theoretical analysis establishes information-theoretic bounds showing non-gamifiability, meaning adversaries cannot improve through training due to fundamental space complexity constraints.

👉 Read the full article on arXiv

OpenAI's new Chief Financial Officer, Sarah Friar, has indicated that the company will focus on 'practical adoption' in 2026.

This strategy aims to make AI technologies more accessible and integrated into daily life and businesses. Friar emphasized the importance of AI becoming a tool that genuinely benefits society, rather than remaining purely experimental.

This focus is part of OpenAI's efforts to reach a broader user base and ensure the ethical and responsible deployment of AI.

👉 Read the full article on The Verge