A Guardian technology report summarises key findings from the 2026 International AI Safety Report, an annual review led by leading AI experts that assesses developments in artificial intelligence and associated risks. The report highlights a number of critical trends shaping the landscape of AI — from the increasing realism of deepfakes and the emergence of emotionally engaging AI companions to cybersecurity implications and labour market uncertainty.
One of the foremost concerns is the rapid spread and realism of deepfakes, particularly AI-generated imagery and pornography that is becoming harder to distinguish from genuine media. This growing sophistication poses risks for misinformation, fraud, and reputational harm, as manipulated content can be deployed in scams, social manipulation, and political interference with increasing reach and impact.
The report also points to the rise of AI companions — conversational agents designed to provide emotional support or social interaction — and notes a trend where some users are forming unusually strong attachments to these systems. While concrete causal evidence is limited, experts are raising concerns about potential mental health effects, emotional reliance and blurred boundaries between artificial and human relationships as these tools become more engaging and prevalent.
Beyond these social issues, the safety review touches on advancements in AI’s abilities — including greater reasoning and scientific problem-solving — alongside ongoing challenges such as vulnerabilities to misuse in cyber-related activities and the uncertain, uneven impact of automation on labour markets. The report underscores that while AI offers significant potential, governing its risks responsibly will be essential to ensure safety, trust and societal well-being as deployment accelerates.
