AI News

Follow the latest AI and automation news

12 Aug 202629 Safar 1448 AH

Scaling AI agents with trustworthy data

Organizations are rapidly adopting agentic AI, yet many struggle to achieve the desired ROI due to inadequate infrastructure. The success of agentic AI hinges on having the right data systems, as it requires seamless access to both structured and unstructured data across the enterprise. Legacy systems often fail to meet these demands, necessitating urgent updates to support real-time decision-making.

Business

12 Aug 202629 Safar 1448 AH

AI tools for breast cancer detection fall short of radiologists' expectations

A survey of 215 members of the Society of Breast Imaging revealed that about half are using FDA-approved AI tools for breast cancer detection. However, the results have not met expectations, with only 35% reporting lower recall rates, while 59% had anticipated such improvements. This gap between promise and reality is evident across all measured categories. This situation highlights significant challenges in the adoption of AI technologies in healthcare, necessitating a reevaluation of the effectiveness of these tools and their impact on patient outcomes.

Models

12 Aug 202629 Safar 1448 AH

DeepSeek V4 Pro Goes Live, Closes In on Frontier Tier

DeepSeek has announced the launch of its V4 Pro model, designated DeepSeek-V4-Pro-0813, featuring a 1 million token input capacity and a 384,000 token output limit. The new model defaults to a thinking mode while also providing access to OpenAI and Anthropic-compatible APIs, allowing teams to switch standards without infrastructure changes. Pricing for the V4 Pro model reflects a clear division from the Flash model, with Pro focusing on complex use cases like coding and agent applications, positioning DeepSeek competitively in the global market with reasonable pricing.

💡

Nvidia Aims to Transform AI Chips into Financial Assets

What impact could this initiative have on the AI market?

Models

MindTopo reveals VLMs’ spatial reasoning abilities

12 Aug 202629 Safar 1448 AH

Models

12 Aug 202629 Safar 1448 AH

MindTopo reveals VLMs’ spatial reasoning abilities

MindTopo has been introduced as a new benchmark for testing topological reasoning in AI, evaluating whether multimodal models can grasp concepts like connectivity, enclosure, and order. This benchmark measures both reasoning and planning, assessing if models can recognize topological relationships in static images and maintain them through a series of actions.

Models

Google's Gemini is losing market share to ChatGPT and Claude according to new market data

12 Aug 202629 Safar 1448 AH

Models

12 Aug 202629 Safar 1448 AH

Google's Gemini is losing market share to ChatGPT and Claude according to new market data

Three data sources indicate that Google's Gemini has lost significant market share in AI, dropping from 12% to 1.9%. In contrast, OpenAI retains over 50% of the market, while Anthropic increased its share from 4.3% to 14.9%. Reports from Pangram, Similarweb, and OpenRouter confirm this trend, reflecting a noticeable shift in user preferences towards ChatGPT and Claude. These figures highlight the intense competition in the AI market, with attention turning to new innovations.

Agents

Google tests AMIE for clinical video consultations

12 Aug 202629 Safar 1448 AH

Agents

12 Aug 202629 Safar 1448 AH

Google tests AMIE for clinical video consultations

Google has launched its medical AI system, AMIE, which conducted video consultations with professional actors and received clinical evaluator ratings comparable to physicians. Fifteen trained actors portrayed various medical conditions, including cardiopulmonary and neurological issues. Google emphasizes that studies involving real patients are essential before clinical use can be established.

Business

NVIDIA CEO Tops Glassdoor’s 2026 List of Best CEOs

12 Aug 202629 Safar 1448 AH

Business

12 Aug 202629 Safar 1448 AH

NVIDIA CEO Tops Glassdoor’s 2026 List of Best CEOs

Jensen Huang, founder and CEO of NVIDIA, has been ranked No. 1 on Glassdoor's Best CEOs list for 2026. He received a 99% approval rating from employees, reflecting his exceptional leadership amid rapid changes in the workplace. This ranking marks the return of the Best CEOs list after a five-year hiatus, replacing the Best-Led Companies list. Glassdoor emphasizes the impact of executive leadership on employee experience and company reputation, with NVIDIA also ranking first in the Best Places to Work in Technology and AI.

Agents

Legal startup founder Robert Mahari joins Anthropic to lead Claude's push into law practices

12 Aug 202629 Safar 1448 AH

Agents

12 Aug 202629 Safar 1448 AH

Legal startup founder Robert Mahari joins Anthropic to lead Claude's push into law practices

Robert Mahari, a legal startup founder, has joined Anthropic as the first "Head of Claude for Legal." He will be responsible for deploying and expanding the Claude AI model within the legal industry. This move comes as interest in AI applications in law grows, with many lawyers seeking to enhance their practice efficiency. Mahari aims to leverage Claude AI to meet the needs of lawyers and their clients. This initiative is expected to transform how lawyers operate, enhancing their ability to provide more effective legal services. It marks a significant shift in the integration of technology within the legal sector.

Governance

Has Anthropic's new watermarking feature opened up another can of AI worms?

12 Aug 202629 Safar 1448 AH

Governance

12 Aug 202629 Safar 1448 AH

Has Anthropic's new watermarking feature opened up another can of AI worms?

Anthropic has launched a new watermarking feature for text generated by its Claude platform, aimed at enhancing transparency and complying with legal requirements. This feature seeks to address challenges posed by technology and the influx of inaccurate content online, yet it raises more questions than it answers. The EU AI Act mandates that users be informed when interacting with AI systems, increasing pressure on tech companies to comply. While firms like OpenAI, Google, and Meta have adapted, Anthropic acknowledges that its new feature has limitations, as a watermark may not conclusively prove that Claude created the content.