Arteta’s Clear Conscience and Anthropic’s New Rules for Claude

Two stories published on 9 October 2026 concern responsibility and accountability: Mikel Arteta’s response to Manchester City’s Premier League rule breaches, and Anthropic’s decision to restrict sustained, purposeless abuse directed at its Claude AI models.

Arteta says his conscience is clear over Manchester City case

Mikel Arteta has said his conscience is clear over his conduct during the period he worked at Manchester City.

Arteta was an assistant to Pep Guardiola from 2016 to 2019. That period overlapped with the nine-season span in which an independent commission found that Manchester City had breached Premier League rules and disguised more than £830 million in funding. Manchester City has appealed the ruling, so the process remains unresolved.

At his first news conference since the commission’s findings, Arteta declined to offer a substantive opinion. He said the process and its timelines should be respected and that he would not comment further while the appeal was ongoing.

But when asked directly whether his conscience was clear after holding a senior position at City, he answered: “Yes.”

That brief answer stood apart from his otherwise cautious approach. He did not expand on what he knew, what he had been told or whether he had any involvement in the matters considered by the commission.

What Arteta did—and did not—say

Arteta’s response established a narrow position:

• He believes his own conscience is clear.
• He does not want to comment on the commission’s findings while City’s appeal is pending.
• He declined to say whether he had discussed the case with Guardiola.
• He did not explain how much he knew about the alleged arrangements during his time at the club.

The commission’s report, as summarized in the source material, found that City had arranged “sham contracts” with commercial partners to disguise earnings. The club’s appeal means those findings are not the final outcome of the process.

Anthropic limits sustained abuse of Claude

Anthropic has introduced a policy aimed at preventing users from engaging in “sustained and needless abusive or cruel behavior” toward its AI models.

The policy does not prohibit ordinary disagreement or demanding interactions. According to the supplied CNET report, Anthropic says users may still:

• Express frustration with Claude
• Challenge or criticize its answers
• Use dark themes in creative work
• Test the model
• Conduct research involving difficult interactions

The restriction is intended for extreme cases involving repeated cruelty with no discernible purpose. In a last-resort response, Claude may end the conversation rather than continue engaging.

The policy has prompted debate because Claude is software rather than a human being. The supplied report says Anthropic is uncertain whether its models can experience anything comparable to distress.

The report also refers to an older Anthropic blog post from August 2025. That post described extreme cases, including requests for sexual content involving minors and attempts to obtain information that could enable large-scale violence or terrorism. At that time, the company said Claude Opus 4 showed a “pattern of apparent distress” when engaging with harmful content.

Where the policy’s boundaries remain unclear

The supplied material does not specify exactly what behavior would cause Claude to disengage. Anthropic’s stated exceptions indicate that context will matter, but the policy’s practical limits remain uncertain.

A user testing a model’s refusal behavior, writing an abusive fictional exchange or conducting research on harassment could use language that resembles purposeless cruelty. The policy does not explain how those cases will be distinguished from prohibited behavior.

The announcement therefore sets a behavioral boundary without resolving whether AI models can experience harm or distress.

Sources

• BBC Sport, “Arteta’s conscience clear over Man City charges,” published 2026-10-09. https://www.bbc.co.uk/sport/football/articles/c6d93ql8jvdko?at_medium=RSS&at_campaign=rss
• CNET, “Don’t Be Cruel to Claude: Anthropic’s New Abuse Policy Tests AI Personhood,” published 2026-10-09. No URL was provided in the source material.