This week brought important news about how much we can trust AI helpers. OpenAI released a new 'Agent' mode for ChatGPT that can act on your behalf. For example, it can order clothes online after checking wedding dress codes and weather. But OpenAI CEO Sam Altman warned this feature is 'cutting edge and experimental'. To keep users safe, they added automatic blocks for dangerous requests (like chemical weapons) and require human approval for big actions. This is the first OpenAI tool labeled 'High Risk' in some areas.

Trust in AI agents is falling worldwide according to a new Capgemini report. Only 27% of people now trust AI agents, down from 43% last year. In the UK, trust dropped from 50% to 32%. The report found only 2% of companies have fully deployed AI agents, though 93% of leaders believe they'll need them to stay competitive. Capgemini's Franck Greverie said companies must 'build trust in AI by ensuring it is developed responsibly'.

Amazon introduced AgentCore, a new set of tools for building safer AI helpers at large scale. At their New York summit, AWS executive Swami Sivasubramanian called this shift 'a tectonic change' in how software works. Amazon is investing $100 million more to develop this technology. They're also starting an AI Agents Marketplace where businesses can find ready-made helpers.

Human roles are changing in this new AI world. People will focus on setting goals, making ethical decisions, and handling creative tasks while AI handles routine work. This could increase human focus on important tasks by 65% and boost job satisfaction by 49%. But big questions remain about who's responsible when AI makes mistakes and how to protect private information.

Europe faces special challenges with these new AI helpers. People in the European Union can't use OpenAI's new agent yet because of upcoming regulations. Some Europeans feel 'over-regulation is making us miss out' on the latest AI tools.

Economic impacts are huge - AI agents could create $450 billion in value within three years. Companies using them well could gain $382 million compared to just $76 million for others. But most companies aren't ready - 80% say their technology systems aren't set up for AI agents yet.

Safety remains crucial as these tools spread. OpenAI published a detailed safety guide for their new agent. Experts stress that humans must stay 'in the loop' to check AI decisions, especially for important tasks. As one tech executive noted, 2025 is the year AI moves 'from answering questions to doing things in the real world'.