Don't blindly trust what AI tells you, says Google's Sundar Pichai (www.bbc.com)

Alphabet CEO Sundar Pichai told the BBC people should not "blindly trust" AI because current models are "prone to errors" and are best used alongside other information tools. His comments come as Google rolls out Gemini 3.0 and an "AI Mode" in Search that integrates its chatbot to provide expert‑style answers — a strategic push to regain ground from rivals like ChatGPT. Pichai framed this as a "new phase" for the AI platform shift, but emphasized the need for a rich information ecosystem and products that are more firmly grounded in accurate data. The remarks matter for developers and users alike because independent BBC testing has already shown significant inaccuracies across major chatbots (OpenAI, Microsoft, Google, Perplexity) when summarizing news — highlighting persistent grounding, verification and hallucination problems. Pichai acknowledged the tension between rapid feature rollout and building robust mitigations, saying Google is increasing investment in AI security proportionally and even open‑sourcing tools to detect AI‑generated images. The takeaway: expect faster, more conversational AI in mainstream products, but also a stronger push toward grounding, detection and safety safeguards — and a continued need for human verification and complementary search or retrieval systems.