IBM Warns on AI Security; Google, D-Wave Advance AI Models & Tools

The landscape of artificial intelligence is rapidly evolving, presenting both unprecedented opportunities for innovation and formidable challenges, particularly in the realm of cybersecurity. Recent developments highlight this dual nature, from IBM’s stark warnings about AI-driven data breaches to groundbreaking collaborations leveraging generative AI for sustainability, and advancements in AI world models and quantum computing toolkits.

IBM’s 2025 Cost of a Data Breach Report casts a critical eye on the burgeoning security risks associated with AI. The report indicates that while global data breach costs saw a slight decrease in 2025, the U.S. experienced a rise, hitting an average of $10.22 million, an all-time high. A significant concern highlighted is the “AI oversight gap,” where organizations are rushing AI adoption without adequate security and governance in place. Alarmingly, 13% of organizations reported breaches related to AI models or applications, with 97% of these lacking basic AI access controls. The report also underscores the growing threat of “shadow AI”—unsanctioned AI tools used within organizations—which can add an average of $670,000 to breach costs. Furthermore, attackers are increasingly weaponizing AI, using it for sophisticated phishing campaigns, deepfake attacks, and even manipulating AI models themselves through techniques like prompt injection and data poisoning. This surge in AI-powered threats necessitates a proactive and agile cybersecurity strategy, emphasizing comprehensive visibility into AI usage and robust data security posture management.

On a more optimistic front, generative AI is proving to be a powerful tool for addressing complex global challenges. Nestlé and IBM Research have partnered to harness generative AI in developing next-generation sustainable food packaging. This collaboration has yielded an AI-powered tool that identifies novel, high-barrier packaging materials capable of enhancing product protection while meeting stringent sustainability, safety, functionality, and cost criteria. By training a chemical language model and using a regression transformer, the AI can design entirely new packaging concepts, significantly shortening the traditionally lengthy research and development cycles. This initiative aligns with Nestlé’s broader AI strategy, which includes optimizing recipes, improving manufacturing efficiency through digital twins, and personalizing nutrition solutions.

Meanwhile, Google is pushing the boundaries of AI simulation with its latest “world model,” Genie 3. This revolutionary AI system can generate diverse, interactive, and physically consistent virtual worlds from simple text prompts, rendering them in real-time at 720p resolution and 24 frames per second. Genie 3 represents a significant leap towards Artificial General Intelligence (AGI) by enabling AI systems to interact with convincing simulations of the real world. This technology holds immense potential for training robots and autonomous vehicles in realistic virtual environments, as well as offering humans immersive experiences for training or exploration.

Further accelerating the pace of AI innovation, D-Wave Quantum Inc. has released an open-source quantum AI toolkit designed to enhance machine learning. This toolkit, part of D-Wave’s Ocean™ software suite, allows developers to seamlessly integrate D-Wave’s annealing quantum computers with PyTorch, a widely used machine learning framework. The toolkit includes a PyTorch neural network module specifically for building and training Restricted Boltzmann Machines (RBMs) using quantum computers, which are crucial for generative AI tasks like image recognition and drug discovery. By offloading computationally intensive RBM training to quantum processors, D-Wave aims to significantly accelerate the speed and reduce the cost of developing new AI models, demonstrating the transformative potential of quantum-enhanced AI.

These concurrent developments underscore a pivotal moment in artificial intelligence. While the industry grapples with the imperative of securing increasingly sophisticated AI systems against evolving threats, the transformative power of AI continues to unlock new possibilities, from sustainable innovation to realistic virtual worlds and the acceleration of complex machine learning tasks through quantum integration. The rapid pace of adoption and innovation means that understanding and addressing AI’s dual nature is more critical than ever.