Microsoft CEO Satya Nadella criticized Anthropic's Fable 5 AI model as 'editorially controlled' during remarks to employees, according to a CNBC report. Microsoft stocks fell 1.3% in overnight trading following the comments, which questioned the AI model's stringent guardrails that route certain user queries to an older version. The criticism comes amid broader debate over AI safety measures as technology companies balance powerful capabilities against potential misuse risks.
MSFT stock is up 7.5% in July, in line with gains across Big Tech as investors rotate back into the sector while selling some of the year's high-flying chip stocks. Investors are closely watching the moves as the second-quarter earnings season kicks off, with Alphabet, Tesla and Intel reporting next week.
Nadella Criticizes Fable 5 Query Routing System
Nadella told engineers working on Microsoft's Copilot AI software that Fable 5's limitations do not 'make sense,' according to a copy of his remarks provided to CNBC. 'If you use Fable, when it refuses for any random thing, it just is like, when was the last time you had a creation tool that was so editorially controlled?' Nadella said.
When users ask Fable about certain topics, including elements of building large-scale AI models, Anthropic may route those queries to an older version of the model, according to Anthropic's support page. An X user highlighted that Fable 5 directed a user to its Opus 4.8 model when asked a deeper question about cancer research. Others also reported seemingly harmless queries being redirected to the older model.
Anthropic Implements Model Access Restrictions
Fable 5 is a scaled-down version of Anthropic's Mythos AI model, which the company has acknowledged is highly capable of identifying vulnerabilities in technology systems and could be misused for hacking. Last month, the U.S. government directed Anthropic to restrict access to Mythos and Fable 5 for foreign users and organizations, prompting the company to temporarily withdraw both models. Anthropic later restored them after briefing officials on the safeguards it had implemented to reduce potential misuse.
Anthropic Plans IPO Investor Meetings
Bloomberg reported earlier this week that Anthropic will meet with investors and banks leading the offering in the coming weeks, and is looking to list its shares as soon as October. OpenAI is moving forward with its IPO plans, but recent reports have said its listing could be pushed to next year.
In April, Anthropic said it tripled its annual revenue run rate to $30 billion, surpassing OpenAI's ARR of about $24 billion. Anthropic is valued at $1.08 trillion, compared to OpenAI's private market valuation of $868.3B, according to data from Nasdaq Private Market.
On Stocktwits, the retail sentiment for MSFT shifted to 'neutral' from 'bearish'. The sentiment was 'neutral' for Anthropic and 'extremely bullish' for OpenAI.
FAQ
What did Satya Nadella say about Anthropic's Fable 5?
Nadella told Microsoft employees that Fable 5 now appears 'editorially controlled' and that some of the model's limits do not 'make sense,' according to a CNBC report.
Why did the U.S. government restrict access to Anthropic's AI models?
Last month, the U.S. government directed Anthropic to restrict access to Mythos and Fable 5 for foreign users and organizations due to the models' capabilities in identifying technology system vulnerabilities that could be misused for hacking.
When is Anthropic planning its IPO?
Bloomberg reported earlier this week that Anthropic will meet with investors and banks in the coming weeks and is looking to list its shares as soon as October.
