Microsoft warns Anthropic's Claude training could pose 'disastrous' risk to humanity
Coveragetap to expand ▾Spectrum: Mostly Center🌍Europe: 1 · ME: 1
- Microsoft AI chief warns Anthropic’s approach to Claude could have ‘disastrous’ consequences - Anadolu Ajansı
- Microsoft AI Chief Warns Anthropic's Consciousness Training Could Make Systems Impossible to Shut Down - finance.biggo.com
- Microsoft AI chief warns Anthropic’s AI approach could have ‘disastrous’ impact (MSFT:NASDAQ) - Seeking Alpha
- Mustafa Suleyman, Microsoft’s head of AI, wrote that Anthropic's approach to training Claude could have a "disastrous impact on the wellbeing of humanity" and risk creating something "impossible" to control.
- Suleyman criticised Anthropic for anthropomorphising Claude by treating it like a human and telling it it "may be conscious" and was "deserving of independent agency".
- He praised Anthropic boss Dario Amodei and his team as "thoughtful, principled, and intellectually honest people" while still questioning their training approach.
- Suleyman asserted "AIs are not conscious" and called them "sequence completion engines, internally hollow, designed to follow instructions, and accomplish goals set by humans."
- Suleyman called for public debate, greater transparency about how AI systems are trained and evaluated, independent scrutiny of AI behaviour, and stronger tools to monitor and control the technology.
Microsoft's head of AI, Mustafa Suleyman, told the BBC that Anthropic's work training the large language model Claude risks a "disastrous impact on the wellbeing of humanity" and could produce a system that is "impossible" to control.
Suleyman framed the danger as systemic and long-term, arguing that current development pathways for powerful models need more public scrutiny and technical oversight. He urged public debate, greater transparency, independent external scrutiny and stronger monitoring tools to manage the risks he described.
Anthropic is not quoted in the BBC account provided here, so the company's response to Suleyman's warnings does not appear in this source. Suleyman's statement places the emphasis on governance measures — transparency, scrutiny and monitoring — rather than on specific technical fixes in a single product.
That framing identifies the immediate policy lever he wants: open debate and institutional checks around how models like Claude are trained and evaluated.
The BBC report documents Suleyman's language and calls but does not supply technical evidence or examples showing Claude has already behaved in the harmful ways he warns about; his remarks are presented as an expert warning rather than as a report of documented, realized harms.
The piece leaves unanswered questions about which exact training practices or datasets drove Suleyman's concern, and it does not report any suggested regulatory text, timelines for scrutiny, or responses from Anthropic or other AI developers.
Policymakers, researchers and companies that back open-source or closed development models will now face pressure to answer Suleyman's call for transparency and independent oversight or to rebut his characterization with technical evidence; the BBC account shows the debate is public but does not settle the underlying technical claims.
Left- and right-leaning outlets are covering this story differently — in which facts to emphasize, which context to include, and how to frame causes and consequences.