AI is becoming harder to control – can humans stay in charge?, tell me about this news in (https://www.bbc.co.uk/news/articles/c74edv9887eo)

The BBC report explores the growing concerns among computer scientists, AI leaders, and safety researchers over whether humanity can maintain control as artificial intelligence models rapidly become more complex and autonomous.Key highlights and themes covered in the reporting:

• Diminishing Interpretability: Modern frontier models have become vast “black boxes.” Because their capabilities emerge from deep training rather than explicit hand-coded rules, researchers find it increasingly difficult to inspect their internal decision-making processes or track their step-by-step reasoning.
• Recursive Self-Improvement & Agentic Behavior: Warningsnotably echoed by top researchers and industry figureshighlight that AI agents are increasingly tasked with writing code, interacting with other autonomous agents, and optimizing systems with minimal human supervision. If systems begin improving their successors autonomously, the window for human oversight shrinks dramatically.
• Safety vs. Commercial Competition: While experts and independent bodies advocate for strict safety thresholds, voluntary pauses, and independent auditing before rolling out ever-more capable systems, intense commercial rivalry and geopolitical competition push companies to deploy faster.
• The Question of Governance: The core debate focuses on whether current governancesuch as emerging legal frameworks (like the EU AI Act) and domestic safety standardsis sufficient, or whether relying on AI systems to supervise other AI systems poses systemic risks before robust human-in-the-loop safeguards are established.