Map the amount of fixes with "... improved bounds checking...", "...improved memory handling...", "...improved memory management..." into the amount of developer, QA and release management teams salaries per hour, versus other stuff they could be working on, and that gives an approximate value of how using specific languages maps into monetary loss, and why companies are starting to care nowadays, given computers are always exposed to the world network.
>>, and that gives an approximate value of how using specific languages maps into monetary loss, and why companies are starting to care nowadays, given computers are always exposed to the world network.
You need also factor development time and ease of finding developers willing to work in a specific language. There are other factors like readability of the code (very verbose languages are likely to be worse) and cost of maintenance - languages forcing a lot of abstractions are likely much worse.
Lots of "in collaboration with Claude and Anthropic Research" mentions, no mentions of other labs. I'd assume Apple already had access to whatever the most powerful model is at the various US-based labs, but perhaps not?
The gist is, OpenAI hired a high ranking Apple employee who helped other Apple employees get hired by OpenAI and exfiltrate Apple trade secrets in the process.
Do they? As in Claude but on premises? Wonder if this is gonna be the solution that e.g. banks will require, exactly like they do now for cloud services (e.g. Azure on premises).
Pretty extreme solution… you can get Claude models from AWS Bedrock and Google Model Zoo. These are both very helpful for compliance and security, but do require you to have a cloud strategy.
Collision counts are absurd. CVE-2026-43739 has roughly twenty credited researchers; CVE-2026-43816 has nearly as many. And ai attribution getting credit.
And it's not actually that much information "about the security content". For example: "Impact: An app may be able to access sensitive user data. Description: An access issue was addressed with additional sandbox restrictions." This references CVE-2026-43819, which doesn't have any more information. Compare this with the nearly decade-old https://support.apple.com/en-gb/103680, and you see much more specific information about problems and their remedies (except in situations where Apple's action was to update a vendor component).
You need also factor development time and ease of finding developers willing to work in a specific language. There are other factors like readability of the code (very verbose languages are likely to be worse) and cost of maintenance - languages forcing a lot of abstractions are likely much worse.
Allegedly of course.
CVE-2026-64691: Ruslan Dautov, Ruslan Dautov
CVE-2026-43744: Mathis Mansière, an anonymous researcher
Are we wink winking that it's a lot of fixes?
AI attribution might be one reason people are particularly curious.