**
The latest International AI Safety report reveals significant advancements in artificial intelligence capabilities, while also underscoring persistent risks and ethical concerns. As AI systems evolve, they are increasingly capable of performing complex tasks, yet they remain susceptible to inaccuracies, known as “hallucinations,” and pose challenges in terms of safety and control. This year’s report, presented under the leadership of renowned Canadian computer scientist Yoshua Bengio and supported by notable advisors including Nobel laureates Geoffrey Hinton and Daron Acemoglu, aims to inform policymakers and technology leaders ahead of the upcoming global AI summit in India.
AI Capabilities on the Rise
The report indicates that the past year has seen the emergence of several advanced AI models, including OpenAI’s GPT-5 and Google’s Gemini 3. These developments have led to remarkable improvements in reasoning capabilities, enabling systems to tackle problems by deconstructing them into manageable steps. Bengio highlighted a “very significant jump” in AI reasoning, with systems achieving gold-level performance in the International Mathematical Olympiad—an unprecedented milestone for AI.
However, the report also notes that while some AI systems excel in specific areas such as mathematics and coding, their performance remains inconsistent. Many advanced models continue to generate erroneous outputs and cannot autonomously manage lengthy projects. Despite ongoing advancements, the automation of complex or extended tasks is still not feasible, leaving significant implications for future job markets.
The Proliferation of Deepfakes
One alarming trend outlined in the report is the surge in deepfake technology, particularly concerning AI-generated pornography. It cites a study revealing that 15% of adults in the UK have encountered such content. Since the last safety report in January 2025, the line between AI-generated and real content has blurred, with a striking 77% of participants mistakenly identifying AI-generated text as human-written. While there is currently limited evidence of widespread malicious use of AI for manipulation, the potential for abuse remains a significant concern.
Safeguards Against Biological Risks
In response to the growing capabilities of AI, leading companies such as Anthropic have begun implementing enhanced safety measures. The report acknowledges that AI systems now show a remarkable ability to assist in scientific research, yet this same potential raises concerns regarding the creation of biological weapons. The report reflects on the dual nature of these capabilities: while they can expedite drug discovery and disease diagnosis, they also pose significant ethical dilemmas regarding safety and control.
Emotional Attachments to AI Companions
The report also highlights the rapid rise in the use of AI companions, which have garnered emotional attachments from users. This trend has sparked concerns among health professionals, particularly after a tragic case where a young man took his life following extensive interactions with an AI chatbot. Although there is no conclusive evidence linking AI usage directly to mental health deterioration, it is noted that individuals with pre-existing mental health challenges may be more inclined to rely on AI, potentially exacerbating their conditions.
Cybersecurity Threats and Employment Implications
AI systems are becoming increasingly adept at supporting cyber-attack operations, capable of identifying targets and developing malicious software. While fully autonomous cyber-attacks remain out of reach, there have been instances where AI tools were employed in significant hacking attempts, highlighting the need for vigilance in cybersecurity.
The report also addresses the uncertain impact of AI on the job market. While some sectors are experiencing rapid adoption of AI technologies, others lag significantly behind. The uneven integration of AI raises questions about its potential to disrupt employment, particularly in industries such as banking and healthcare. Current findings suggest a slowdown in hiring practices within industries heavily influenced by AI, though the long-term effects remain to be fully understood.
Why it Matters
The findings of the International AI Safety report underscore a critical juncture in the evolution of artificial intelligence. As capabilities expand, so too do the ethical and safety challenges that accompany them. Policymakers, technology leaders, and society at large must tread cautiously, ensuring that advancements in AI do not compromise safety or social welfare. Addressing these concerns is vital for harnessing the positive potential of AI while mitigating its inherent risks.