“Don’t blindly trust AI”: Sundar Pichai issues strong warning as Google faces criticism over Gemini errors

Alphabet CEO Sundar Pichai has delivered one of his clearest public warnings yet: “People should not blindly trust everything AI tells them.”
In an exclusive BBC interview, the Google chief acknowledged that today’s AI models — including Google’s own systems — are “prone to errors” and must be used with caution.

Pichai’s statement comes after Google’s AI Overviews feature faced widespread ridicule for generating inaccurate and sometimes bizarre summaries of search results.

“We take pride in trying to provide the most accurate information possible, but the current state-of-the-art AI technology is prone to errors,” he said.
“People have to learn to use these tools for what they’re good at — not blindly trust everything they say.”

Experts: “AI makes things up to please us — that’s dangerous”

Professor Gina Neff, a leading expert in responsible AI at Queen Mary University of London, issued her own warning:

“We know these systems make up answers to please us — and that’s a problem.”

She stressed that the risks escalate dramatically when AI is used for:

• health information,

• mental well-being,

• science,

• or news.

Her sharpest criticism targeted big tech directly:

“The company is asking us to mark their own exam paper while they’re burning down the school.”

Google admits limitations but highlights efforts

Pichai argued that this is precisely why Google maintains a broader ecosystem of information tools:

“That’s why people also use Google Search — we have more grounded products designed to provide accurate information.”

Google displays disclaimers on all AI tools, warning they may make mistakes.
But that hasn’t shielded the firm from backlash — especially after BBC research earlier this year found “significant inaccuracies” in responses generated by:

• OpenAI’s ChatGPT,

• Microsoft’s Copilot,

• Google’s Gemini,

• and Perplexity AI.

Gemini 3.0: Google’s comeback weapon

The tech world has been eagerly awaiting Gemini 3.0, Google’s newest consumer AI model that has started reclaiming market share from ChatGPT.
Google’s new “AI Mode” in Search integrates Gemini directly into results, aiming to simulate talking to an expert.

But rapid innovation brings risks.

“There is tension between how fast the technology is moving and how fast we can build mitigations,” Pichai admitted.

Alphabet expands AI safety alongside AI power

Pichai said Alphabet has increased AI safety investment proportionally to AI development itself.
Measures include:

• open-sourcing tools that detect AI-generated images,

• boosting systems to prevent misinformation,

• enhancing internal validation pipelines.

Asked about Elon Musk’s DeepMind “dictatorship” fears

Pichai responded to resurfaced comments from Elon Musk, who once warned OpenAI founders that DeepMind could lead to an AI “dictatorship”.

His answer:

“No one company should own a technology as powerful as AI.”

But he added that the current landscape includes many players:
“We are far from a world where one company builds all AI and everyone must use it.”

The message Pichai wants everyone to hear

“Use AI wisely. Check facts. Don’t trust everything it says.”

As the AI revolution accelerates, the challenge is no longer merely building powerful models — but ensuring humanity keeps control.