openai, google and apple Updates

OpenAI recently rolled out its new AI image generation model, Image Model 1.5, making it available to all ChatGPT users globally and through its API as GPT-Image-1.5. Launched on December 16, 2025, this model is four times faster than its predecessor and offers enhanced editing capabilities, allowing users to easily add, change, or remove objects. It also significantly improves the readability of text within AI-generated images, a common hurdle for such technology. OpenAI's CEO of applications, Fidji Simo, highlighted the goal of making AI tools more useful, noting that a dedicated space in ChatGPT's sidebar simplifies image creation. This new model directly competes with offerings like Google's Imagen 2. Beyond creative applications, AI continues to reshape various sectors. A McKinsey Global Institute study indicates that while AI will handle routine digital tasks, human skills will remain vital, shifting focus to unique abilities like judgment and questioning. Demand for skills to manage AI tools has surged sevenfold, with AI expected to augment human work in fields like radiology and customer service. McKinsey projects AI could contribute nearly $2.9 trillion to the US economy by 2030. Meanwhile, health AI experts, as reviewed by Brittany Trang from STAT's AI Prognosis newsletter, are assessing their 2025 predictions, with a look ahead to 2026. In finance, Julia Khandoshko, CEO of Mind Money, anticipates that by 2026, AI's role will solidify as an assistant, with human oversight remaining crucial for independent decision-making. The crypto industry is emerging as a core technology, expanding investment opportunities, including access to assets like Apple shares through digital derivatives. This digitalization also intensifies cybersecurity challenges, demanding proactive development. ESG standards continue as a regulatory requirement, with financial institutions now needing to report environmental impacts. Security remains a critical area for AI application. Amazon's Chief Security Officer, Steve Schmidt, reported a 27 percent increase in North Korean IT worker applications in 2025, with some individuals charged for stealing millions. Amazon employs AI tools alongside human efforts to detect unusual contact information or fake company histories, requiring more in-person interviews and using its Midway authentication system with physical security keys. AI also helps Amazon speed up security analysis, detect fake reviews, and identify flaws in AI-written code. Similarly, Cisco announced on December 17, 2025, it will use its specialized AI model, "Foundation-Sec-1.1-8B-Instruct," to power security products like Duo Identity Intelligence, providing actionable insights into potential identity issues for over 2,000 customers. The demand for AI is also fueling significant growth in the semiconductor industry. The global market for semiconductor equipment expects major expansion through 2027, driven by AI applications. Countries like Taiwan, South Korea, and China are heavily investing in advanced technologies such as High Bandwidth Memory (HBM) and the new 2-nanometer process node, essential for next-generation AI hardware. This growth is prompting companies to upgrade factories and acquire new machinery for wafer fabrication, lithography, etching, and testing. Geopolitically, President Donald Trump approved previously banned exports of Nvidia's H200 AI chips to China on December 17, 2025, a move that could bolster the People's Liberation Army's technological capabilities, particularly in military AI applications like drone swarms. Experts warn this decision risks narrowing America's military advantage. Finally, efforts to make AI more energy-efficient are advancing. On December 17, 2025, a team led by Tianyi Chen from Cornell Tech, in collaboration with IBM and Rensselaer Polytechnic Institute, unveiled a new analog computing method. Their "Residual Learning" technique corrects hardware imperfections in real time, enabling analog in-memory computing (AIMC) chips to train AI models with accuracy comparable to digital systems but using significantly less energy. This breakthrough could make large-scale AI training more affordable and expand its applications from edge devices to data centers.

Key Takeaways

• OpenAI launched its GPT Image 1.5 model on December 16, 2025, offering 4x faster image generation, improved editing tools, and better text rendering, available to all ChatGPT users and via API.
• The McKinsey Global Institute predicts AI could add nearly $2.9 trillion to the US economy by 2030, with human skills adapting to focus on judgment and questioning, and a sevenfold increase in demand for AI management skills.
• The global semiconductor equipment market is projected for significant growth through 2027, driven by AI demand, with countries like Taiwan, South Korea, and China investing in HBM and 2-nanometer process nodes.
• President Donald Trump approved exports of Nvidia's H200 AI chips to China on December 17, 2025, raising concerns about strengthening the People's Liberation Army's military AI capabilities.
• Amazon's Chief Security Officer, Steve Schmidt, reported a 27% rise in North Korean IT worker infiltration attempts in 2025, with Amazon using AI and human efforts, including its Midway authentication system, to combat these threats.
• Cisco introduced its specialized AI model, "Foundation-Sec-1.1-8B-Instruct," on December 17, 2025, to power cybersecurity products like Duo Identity Intelligence, providing actionable insights for over 2,000 customers.
• Researchers from Cornell Tech, IBM, and Rensselaer Polytechnic Institute developed a "Residual Learning" method for analog in-memory computing (AIMC) on December 17, 2025, significantly reducing energy consumption for AI model training while maintaining accuracy.
• By 2026, AI is expected to solidify its role as an assistant in finance, with human responsibility for independent decision-making, while the crypto industry expands investment opportunities, including access to Apple shares through digital derivatives.
• Health AI experts reviewed their 2025 predictions, as detailed in Brittany Trang's STAT AI Prognosis newsletter, assessing advancements in the field.
• OpenAI's new GPT Image 1.5 model directly competes with existing offerings like Google's Imagen 2 in the AI image generation space.

OpenAI launches faster AI image model 1.5 in ChatGPT

OpenAI released its new AI image generation model, Image Model 1.5, on December 16, 2025. This model is now available to all ChatGPT users globally and through its API as GPT-Image-1.5. It is four times faster than the previous version and offers improved editing tools, allowing users to add, change, or remove objects. The new model also creates more readable text in AI-generated images, a common challenge for such technology. Fidji Simo, OpenAI's CEO of applications, noted that a dedicated space in ChatGPT's sidebar makes image creation easier.

OpenAI's new image tool makes photo editing simple

OpenAI launched its new ChatGPT image generator, GPT Image 1.5, on Tuesday, making photo editing much easier. This "native multimodal" model allows users to alter images by simply typing sentences, changing poses, removing objects, or adjusting styles. It can also keep a person's face consistent across different edits. Fidji Simo, OpenAI's CEO of applications, stated the goal is to make AI tools more useful. While it competes with Google's Imagen 2, the model includes filters to block harmful content and shows improved text rendering in images.

Workers will adapt and thrive in the AI era

A McKinsey Global Institute study shows that most human skills will remain useful as AI advances, but how they are applied will change. AI will handle routine digital tasks, allowing people to focus on unique human abilities like asking questions and exercising judgment. There is a sevenfold increase in demand for skills to use and manage AI tools. Examples like radiology and customer service show AI augmenting human work, not replacing it. Leaders will need AI fluency to guide human-AI teams, and McKinsey estimates AI could add nearly 2.9 trillion dollars to the US economy by 2030.

Health AI experts review 2025 predictions

Brittany Trang, Ph.D., from STAT's AI Prognosis newsletter, published an article on December 17, 2025. She asked health AI experts to review their "bingo cards" of predictions for the year 2025. The article looks back at whether these experts were correct about the advancements in health artificial intelligence. The newsletter will return on January 7 with predictions for 2026.

AI demand boosts semiconductor equipment sales until 2027

The global market for semiconductor equipment expects major growth through 2027, driven by the increasing demand for AI applications. Countries like Taiwan, South Korea, and China are investing heavily in advanced semiconductor technologies. These include High Bandwidth Memory HBM and the new 2-nanometer process node, which are crucial for powering next-generation AI hardware. Companies are upgrading their factories and buying new machinery to meet the need for more powerful and specialized AI chips. This growth will continue to fuel demand for equipment used in wafer fabrication, lithography, etching, and testing.

AI and crypto reshape finance in 2026

Julia Khandoshko, CEO of Mind Money, believes that by 2026, AI will need to define its role in finance. While AI is useful as an assistant, its use for independent decision-making faces skepticism, requiring a human to be responsible for final choices. The crypto industry is becoming a core technology, offering new investment opportunities and expanding access to assets like Apple shares through digital derivatives. This digitalization also brings major cybersecurity challenges, requiring proactive development. ESG standards remain important as a regulatory requirement, with banks and fintechs now needing to report their environmental impacts.

Amazon fights North Korean IT worker infiltration with AI

Steve Schmidt, Amazon's Chief Security Officer, revealed on December 17, 2025, that North Korean IT workers are increasingly trying to infiltrate the company. Amazon saw a 27 percent rise in such applications in 2025, with some individuals charged for stealing millions. These attackers now buy legitimate American identities to get hired. Amazon uses a mix of AI tools and human efforts to fight back, training AI to spot unusual contact info or fake company histories. The company also requires more in-person interviews and uses its Midway authentication system with physical security keys. AI helps Amazon speed up security analysis, detect fake reviews, and find flaws in AI-written code.

Nvidia chips could boost China's military AI power

President Donald Trump approved previously banned exports of Nvidia's H200 AI chips to China, a move that could strengthen the People's Liberation Army's technological goals. Columnist Karishma Vaswani reported on December 17, 2025, that this decision risks narrowing America's military advantage, especially in the Indo-Pacific region. The H200 is Nvidia's second-most powerful AI processor, and the PLA seeks advanced chips for military AI applications like drone swarms. While Chinese regulators are still reviewing the deal, experts warn that allowing access could accelerate Beijing's military modernization. US policymakers face a difficult choice between restricting exports, which might push China to develop its own chips faster, or allowing access, which could boost its military AI capabilities.

New analog method cuts AI training energy use

Researchers have developed a new analog computing method that significantly reduces the energy needed to train AI models. On December 17, 2025, a team led by Tianyi Chen from Cornell Tech, with IBM and Rensselaer Polytechnic Institute, presented their work on analog in-memory computing AIMC. While analog chips are energy-efficient for running trained AI, training them was difficult due to hardware imperfections. Their "Residual Learning" method corrects these errors in real time, allowing analog chips to train AI models with accuracy similar to digital systems but using much less energy. This breakthrough could make large-scale AI training more affordable and enable new applications, from edge devices to data centers.

Cisco uses its own AI model for security products

On December 17, 2025, Cisco announced it will use its own AI model, "Foundation-Sec-1.1-8B-Instruct," to power its products, starting with Duo Identity Intelligence. This model is specifically tuned for cybersecurity tasks like automating security operations, defending against threats, and helping engineering teams. It examines user activity after login to spot unusual patterns, such as strange locations or suspicious privilege use. The AI will create weekly email digests for over 2,000 Cisco customers, providing clearer and more actionable insights into potential identity issues. Cisco believes its specialized model offers better accuracy and relevance than general-purpose AI.

Sources

• OpenAI Just Dropped a New AI Image Model in ChatGPT to Rival Google's Nano Banana
• OpenAI’s new ChatGPT image generator makes faking photos easy
• How Workers Will Adapt in the AI Era
• AI Prognosis: Were experts right about their 2025 health AI ‘bingo cards’?
• AI Demand Drives Semiconductor Equipment Sales Growth Through 2027
• The Shape of Finance in 2026: How AI, ESG, and Global Risk Are Redefining Banking and Investment Strategy
• How Amazon’s CSO defends against efforts by North Korean IT workers to infiltrate his company
• Will Nvidia Turbocharge China’s AI Military?
• New analog computing method slashes AI training energy use
• Cisco decides homegrown AI model is fit to power its wares