Anthropic pricing and security wobble
What happened
- Anthropic reportedly tested removing Claude Code from its Pro plan and briefly altered public pricing documentation. - Security researchers and cyber experts warned Claude models have produced vulnerable code in some instances. - The episode raises vendor trust, pricing predictability, and code-security questions for teams adopting AI coding tools. ( )
Why it matters
Anthropic briefly tested removing Claude Code from its $20-a-month Pro plan this week, then said the change applied to about 2% of new signups. (arstechnica.com) (theregister.com) Public pages changed at the same time. The Register reported that Anthropic’s pricing page dropped the line saying Pro “includes Claude Code,” and showed an “X” for Claude Code on Pro, even as other Anthropic pages still said Pro users had access. (theregister.com) (anthropic.com) Anthropic head of growth Amol Avasare said the company was “running a small test” on about 2% of new prosumer signups and that existing Pro and Max subscribers were not affected. He said Max was built for heavy chat use before Claude Code, Cowork, and long-running agents became common workflows. (theregister.com) (arstechnica.com) Claude Code is not a simple autocomplete tool. Anthropic describes it as an “agentic coding system” that can read a codebase, edit files across a project, run tests, and deliver committed code. (anthropic.com) That makes pricing changes more sensitive for small teams that build workflows around one subscription tier. Anthropic has also been pitching Claude Code to larger organizations with spend caps, seat controls, and usage analytics meant to make costs more predictable. (anthropic.com 1) (anthropic.com 2) At the same time, security researchers told Forbes that recent Claude models have generated code with serious weaknesses in some tests. The report said experts saw a drop in secure coding performance after a model update, though Anthropic had not publicly explained the change. (forbes.com) Anthropic has been marketing Claude for the opposite job too. On February 20, the company announced Claude Code Security, a research-preview tool for Team and Enterprise customers that scans codebases for vulnerabilities and suggests patches for human review. (anthropic.com) In that announcement, Anthropic said Claude can help find subtle flaws that rule-based scanners miss, but also warned that the same capabilities could help attackers exploit software. The company said nothing is applied automatically and developers must approve every fix. (anthropic.com) Anthropic has recently kept expanding Claude’s coding pitch. Its April 16 launch post for Claude Opus 4.7 said the model improved on hard software-engineering tasks, while the Claude Code product page says most code at Anthropic is now written by Claude Code. (anthropic.com 1) (anthropic.com 2) For customers, the week’s problem was not one bug or one pricing test by itself. It was a coding product sold as central to software work appearing to shift plans in public while outside researchers questioned how safely its latest models write code. (arstechnica.com) (forbes.com) (anthropic.com)
Key numbers
- ( ) Anthropic briefly tested removing Claude Code from its $20-a-month Pro plan this week, then said the change applied to about 2% of new signups.
- (theregister.com) (anthropic.com) Anthropic head of growth Amol Avasare said the company was “running a small test” on about 2% of new prosumer signups and that existing Pro and Max subscribers were not affected.
- (anthropic.com 1) (anthropic.com 2) At the same time, security researchers told Forbes that recent Claude models have generated code with serious weaknesses in some tests.
- On February 20, the company announced Claude Code Security, a research-preview tool for Team and Enterprise customers that scans codebases for vulnerabilities and suggests patches for human review.
What happens next
- Anthropic briefly tested removing Claude Code from its $20-a-month Pro plan this week, then said the change applied to about 2% of new signups.
- (anthropic.com) In that announcement, Anthropic said Claude can help find subtle flaws that rule-based scanners miss, but also warned that the same capabilities could help attackers exploit software.
- Its April 16 launch post for Claude Opus 4.7 said the model improved on hard software-engineering tasks, while the Claude Code product page says most code at Anthropic is now written by Claude Code.
Quick answers
What happened in Anthropic pricing and security wobble?
Anthropic reportedly tested removing Claude Code from its Pro plan and briefly altered public pricing documentation. Security researchers and cyber experts warned Claude models have produced vulnerable code in some instances. The episode raises vendor trust, pricing predictability, and code-security questions for teams adopting AI coding tools. ( )
Why does Anthropic pricing and security wobble matter?
Anthropic briefly tested removing Claude Code from its $20-a-month Pro plan this week, then said the change applied to about 2% of new signups. (arstechnica.com) (theregister.com) Public pages changed at the same time. The Register reported that Anthropic’s pricing page dropped the line saying Pro “includes Claude Code,” and showed an “X” for Claude Code on Pro, even as other Anthropic pages still said Pro users had access. (theregister.com) (anthropic.com) Anthropic head of growth Amol Avasare said the company was “running a small test” on about 2% of new prosumer signups and that existing Pro and Max subscribers were not affected. He said Max was built for heavy chat use before Claude Code, Cowork, and long-running agents became common workflows. (theregister.com) (arstechnica.com) Claude Code is not a simple autocomplete tool. Anthropic describes it as an “agentic coding system” that can read a codebase, edit files across a project, run tests, and deliver committed code. (anthropic.com) That makes pricing changes more sensitive for small teams that build workflows around one subscription tier. Anthropic has also been pitching Claude Code to larger organizations with spend caps, seat controls, and usage analytics meant to make costs more predictable. (anthropic.com 1) (anthropic.com 2) At the same time, security researchers told Forbes that recent Claude models have generated code with serious weaknesses in some tests. The report said experts saw a drop in secure coding performance after a model update, though Anthropic had not publicly explained the change. (forbes.com) Anthropic has been marketing Claude for the opposite job too. On February 20, the company announced Claude Code Security, a research-preview tool for Team and Enterprise customers that scans codebases for vulnerabilities and suggests patches for human review. (anthropic.com) In that announcement, Anthropic said Claude can help find subtle flaws that rule-based scanners miss, but also warned that the same capabilities could help attackers exploit software. The company said nothing is applied automatically and developers must approve every fix. (anthropic.com) Anthropic has recently kept expanding Claude’s coding pitch. Its April 16 launch post for Claude Opus 4.7 said the model improved on hard software-engineering tasks, while the Claude Code product page says most code at Anthropic is now written by Claude Code. (anthropic.com 1) (anthropic.com 2) For customers, the week’s problem was not one bug or one pricing test by itself. It was a coding product sold as central to software work appearing to shift plans in public while outside researchers questioned how safely its latest models write code. (arstechnica.com) (forbes.com) (anthropic.com)
