ChatGPT Agent for ‘deep research’ 🔍, Microsoft Employees Terminations ❌, Figma AI 🖌️

In partnership with

Hire an AI BDR to Automate Your LinkedIn Outreach

Sales reps are wasting time on manual LinkedIn outreach. Our AI BDR Ava fully automates personalized LinkedIn outreach using your team’s profiles—getting you leads on autopilot.

She operates within the Artisan platform, which consolidates every tool you need for outbound:

• 300M+ High-Quality B2B Prospects

• Automated Lead Enrichment With 10+ Data Sources Included

• Full Email Deliverability Management

• Personalization Waterfall using LinkedIn, Twitter, Web Scraping & More

Book a demo to see what Ava can do.

💬 OpenAI Developments

OpenAI Unveils A.I. Tool That Can Do Research Online

OpenAI has introduced "Deep Research," an AI agent powered by their o3 model, designed to conduct extensive web browsing, data analysis, and synthesis, delivering comprehensive reports within 5 to 30 minutes. This tool aims to enhance research efficiency by automating complex information gathering and analysis tasks.

OpenAI Strikes Deal With US Government to Use Its AI for Nuclear Weapon Security

OpenAI has partnered with the U.S. National Laboratories to integrate its o1 series AI models into nuclear security initiatives, aiming to reduce the risk of nuclear war and secure nuclear materials and weapons worldwide. However, concerns have been raised about the reliability of these AI models, citing instances of data leaks and generating false information.

OpenAI, SoftBank Form Joint Venture for AI Services

OpenAI and SoftBank Group have established a joint venture named SP OpenAI Japan to expand AI services for businesses globally, starting with Japanese companies. The venture will deploy an advanced AI system called "Cristal intelligence" within SoftBank's subsidiaries, including Arm Holdings, aiming to enhance productivity, with SoftBank committing to invest $3 billion annually in OpenAI's technology across its operations.

🌐 AI in Robotics

Watch China's 4-legged 'Black Panther 2.0' robot run as fast as Usain Bolt

Researchers from Mirror Me and Zhejiang University's Center for Interdisciplinary Mechanics in China have developed "Black Panther 2.0," a quadrupedal robot capable of sprinting 100 meters in under 10 seconds, reaching speeds up to 10.4 meters per second (23.3 mph). This performance is achieved through flexible, knee-like joints and reinforced carbon-fiber structures, allowing the robot to approach the world-record pace set by Usain Bolt in 2009.

Barcelona Airport launches the world’s first AI-powered robotic restaurant to enhance passenger dining experience

Barcelona Airport has unveiled SELF, the world's first AI-powered robotic restaurant in an airport, in collaboration with Areas, a Spanish multinational in food & beverage and travel retail. Covering 137 square meters, SELF offers a variety of fresh, high-quality items such as coffee, premium sandwiches, juices, pastries, salads, pokes, desserts, and soft drinks. The restaurant features an advanced robotic arm capable of handling six orders simultaneously, with customers placing orders via touchscreen and receiving notifications when their tray is ready. AI-powered machine vision optimizes real-time decisions, enhancing speed and service quality.

🚀 Tech Industry Moves

DeepSeek's cost-effective AI models are enabling European tech startups to adopt advanced AI technology more readily, potentially narrowing the gap with U.S. competitors. Companies like Germany's Novo AI have transitioned from OpenAI's ChatGPT to DeepSeek, citing significant cost savings without compromising performance.

DeepSeek: Recent evaluations by Cisco and the University of Pennsylvania revealed that DeepSeek's R1 AI model failed to block any of the 50 harmful prompts tested, resulting in a 100% attack success rate. This raises significant concerns about the model's safety measures, especially when compared to other leading AI models that demonstrate at least partial resistance to such attacks.

Nvidia: Helen Toner argues that maintaining U.S. export restrictions on Nvidia’s advanced AI chips is crucial because China’s AI advancements, particularly through companies like DeepSeek, could gain a competitive edge if given access to top-tier hardware. These chips are essential for training and running large-scale AI models, and lifting restrictions would accelerate China’s AI development in critical sectors like military applications, cybersecurity, and autonomous systems. By keeping restrictions in place, the U.S. aims to limit China’s ability to surpass Western AI capabilities and maintain geopolitical and technological advantages.

Nvidia has responded to AMD's recent benchmarks by asserting that its RTX 4090 GPU outperforms AMD's RX 7900 XTX by nearly 50% in DeepSeek AI model tests. In these evaluations, Nvidia claims the RTX 5090 is up to 2.2 times faster than the RX 7900 XTX, while the RTX 4090 is approximately 47% faster, challenging AMD's earlier assertions of superior performance.

Anthropic has developed "constitutional classifiers," a system designed to prevent large language models from generating harmful content by monitoring both inputs and outputs. In tests, this approach enabled their Claude 3.5 Sonnet model to reject over 95% of harmful prompts, significantly improving safety measures against "jailbreaking" attempts.

Meta has officially launched its AI assistant, Meta AI, in Iraq and the broader Middle East and North Africa region, integrating it across Facebook, Instagram, WhatsApp, and Messenger. Powered by the Llama 3.2 large language model, Meta AI supports Arabic and includes safeguards to ensure accuracy, prevent harmful content, and enhance user experiences with features like AI-generated images, content recommendations, and chatbot interactions.

Microsoft has initiated immediate terminations for employees deemed to have underperformed, revoking their access to company systems and facilities without offering severance packages. These actions are part of a broader strategy to address workforce performance, with the company emphasizing its commitment to high performance and appropriate action when standards are not met.

Microsoft: Mobiliar, Switzerland's oldest private insurance company, has enhanced its customer service by integrating Microsoft's Azure OpenAI Service into its operations. The company utilizes Mobi-ChatGPT to automate email processing, generate text, perform translations, and analyze data, thereby streamlining processes and improving customer experiences.

Microsoft has unveiled Microsoft 365 Copilot Chat, a pay-as-you-go service that integrates AI agents into its existing free chat tool for educational institutions. These generative AI-powered agents can autonomously perform tasks tailored to specific educational needs, such as instructional design and providing insights based on institutional data. The service includes secure AI chat powered by GPT-4o, customizable agents, and IT controls for data protection and management. Institutions like Florida State University are already utilizing Copilot Chat to enhance various educational processes, including delivering course information and facilitating data analysis for researchers.

⚖️ AI Ethics and Regulation

The EU AI Act has officially begun enforcement, banning AI systems deemed to pose "unacceptable risk", such as social scoring, biometric surveillance, and emotion recognition in workplaces and schools. Companies violating these rules face fines of up to €35 million or 7% of global revenue. While most tech firms pledged early compliance, enforcement won’t take full effect until August 2025. Exemptions exist for law enforcement and medical use, but uncertainty remains regarding how the AI Act will interact with existing regulations like GDPR.

💡 Emerging AI Technologies and Companies

MLCommons and Hugging Face team up to release massive speech dataset for AI research

MLCommons and Hugging Face have released Unsupervised People’s Speech, one of the largest public domain speech datasets, featuring over a million hours of audio in 89+ languages to advance AI speech research. While the dataset aims to improve speech recognition and synthesis, concerns remain over bias—most recordings are in American-accented English—and ethical issues, as some contributors may be unaware their voices are used for AI training. MLCommons pledges to refine the dataset, but researchers are urged to exercise caution when using it.

Google’s X spins out Heritable Agriculture, a startup using AI to improve crop yield

Google’s X moonshot factory has spun out Heritable Agriculture, a startup using AI and machine learning to optimize crop breeding for higher yields, lower water use, and increased carbon storage. Founded by Brad Zamft, the company analyzes plant genomes to improve agriculture without genetic modification for now. Heritable has tested its models in controlled chambers and field sites and has secured funding from Google, FTW Ventures, Mythos Ventures, and SVG Ventures. The startup aims to modernize the 12,000-year-old farming industry through data-driven breeding techniques.

🎨 AI in Creative Industries

Figma has introduced new AI-powered features in its beta version, including "Boost Resolution" and "Remove Background" tools to enhance image quality directly in the editor. The Assets Community search now allows users to find design components using keywords or image uploads, improving accessibility. Additionally, Figma has updated DevMode with gradient variable binding, enabling better design-to-development workflow consistency. These features are rolling out gradually as part of Figma AI’s limited beta.

If you want more updates related to AI, subscribe to our Newsletter