Yoshua Bengio, widely regarded as one of the founding figures of modern artificial intelligence, has long been a leading voice in machine learning. But in recent years, his perspective on the technology has evolved, driven by growing concerns about its real-world risks. In a recent conversation, Bengio shared his deepening unease about the current trajectory of AI development, stating plainly: “Malicious use is already happening.” Once focused on advancing the science of learning algorithms, Bengio now emphasizes the urgent need to build safety into AI systems from the ground up. He argues that the field must shift from simply creating more powerful models to ensuring they are aligned with human values and resistant to abuse. Bengio’s concerns are rooted in observable trends. From deepfakes used in disinformation campaigns to AI-driven cyberattacks and automated fraud, the tools are already being weaponized. “We’re not just talking about hypothetical future dangers,” he said. “We’re seeing harmful applications unfold in real time.” To counter these threats, Bengio is leading efforts to develop AI systems that are inherently safer—what he calls “AI with safety built in.” This includes research into techniques like interpretability, robustness against adversarial inputs, and mechanisms to prevent models from generating harmful content. He also advocates for stronger regulatory frameworks, ethical guidelines, and international cooperation to govern AI development. Despite the risks, Bengio remains hopeful. He believes that with the right focus, AI can be a force for good—improving healthcare, accelerating scientific discovery, and solving complex global challenges. But he stresses that this potential can only be realized if safety and ethics are prioritized at every stage of development. “AI is not inherently good or bad,” he said. “It’s a tool. But like any powerful tool, it demands responsibility. The question isn’t whether we can build smarter AI—it’s whether we can build it wisely.” Bengio’s call to action reflects a broader shift in the AI community, where researchers, policymakers, and industry leaders are increasingly recognizing that innovation must go hand in hand with accountability. As AI continues to evolve at breakneck speed, his message is clear: the time to act is now.