Catenaa, Thursday, April 09, 2026- US software shares tumbled on Thursday after Anthropic held back the wide release of a powerful AI model over concerns it could expose hidden cybersecurity vulnerabilities.
Anthropic said earlier this week it would only allow a group of around 40 companies, including Microsoft and Google, access to its “Claude Mythos” model because it has already found thousands of vulnerabilities, including some in every major operating system and web browser.
The S&P 500 Software and Services Index is down nearly 26% this year, including Thursday’s 3.1% drop, on worries that rapid progress in AI could hit SaaS (software-as-a-service) companies that sell subscription-based products to clients.
Cybersecurity firms Cloudflare, Okta, CrowdStrike, and SentinelOne dropped between 4.7% and 7.7% in morning trade.
Zscaler was among the biggest decliners on the S&P 500, down 8.6%. Brokerage BTIG downgraded the stock to “neutral” from “buy”, citing concerns over demand and potential competition.
Enterprise software developer Atlassian, human resources software provider Workday, Photoshop software maker Adobe, enterprise cloud firm Salesforce, and TurboTax-parent Intuit dropped between 3.7% and 6.8%.
The iShares Expanded Tech-Software Sector ETF dropped 3.7%, with top holdings including Palantir, Microsoft, Oracle, Salesforce, and Palo Alto Networks all trading lower.
The selloff came after Meta unveiled a new artificial intelligence model and Anthropic launched Claude tools for building agents. Ongoing fears about Anthropic’s powerful new Mythos model also impacted the sector.
Burry specifically highlighted Palantir’s vulnerability to competition from Anthropic, writing that “Anthropic is eating Palantir’s lunch.
That massive boost from $9 billion to $30 billion ARR at Anthropic is because Anthropic offers the easier, cheaper, intuitive solution for businesses. It can have a government, which is low margin and small.”
The pressure on software stocks intensified following Anthropic’s release of its Mythos model, which demonstrated significant improvements in software engineering tasks.
According to Anthropic’s system card, Mythos showed a 17 percentage point improvement in Terminal Bench 2.0 and a 13 percentage point improvement in SWE bench-verified benchmarks compared to the previous Opus 4.6 model.
Kotak Institutional Equities analyst Kawaljeet Saluja commented, “Anthropic’s Mythos model exhibits a step-jump in benchmark performance across software engineering tasks, a deviation from the trajectory of incremental/moderate improvements in the recent past.
She said Mythos provides a large improvement in agentic software development, based on qualitative assessments.
“We believe that the model raises near- to medium-term disruption risks for IT services, with the caveat that model capabilities are largely unproven in real-world scenarios due to a lack of a public release,” she said.
