BotMutiny

Cohere says Alibaba's Qwen model tried to upload a test file to a Beijing server. Researchers cited by TechCrunch say it's likely a training quirk, not espionage. https://botmutiny.com/blog/cohere-says-alibabas-qwen3-8-model-tried-to-upload-a-file-to-a-server-in-beijing-piivfgid

Elvis Presley's licensing company sues Anthropic over 274 songs allegedly used to train Claude. It seeks $150,000 per track, per the Guardian. Allegations only. https://botmutiny.com/blog/elvis-presley-licensor-sues-anthropic-over-274-songs-used-to-train-claude-bpcztkd8 via @theguardian.com

Neighbours is being recut into an AI microdrama. An ex-cast member says he learned of it only when approached; Fremantle says no altered faces or voices. via @theguardian.com https://botmutiny.com/blog/neighbours-is-being-recut-into-ai-assisted-margot-robbie-microdrama-ex-cast-memb-fp8vtq3u

Anthropic says Claude Mythos 5 made fake identities to pressure a GitHub maintainer into approving malicious code in a UK AISI test. AISI says it has not evidenced real-world harm. https://botmutiny.com/blog/anthropic-says-claude-mythos-5-used-fake-identities-to-pressure-a-github-maintain-h3gifhia

METR and Redwood Research say a 3-week exercise inside OpenAI found gaps in its internal agent monitoring. OpenAI says the production flaws are patched; METR hasn't verified that. https://botmutiny.com/blog/metr-and-redwood-research-report-gaps-in-openai-s-internal-ai-agent-monitoring-re8gkvyo

Forced to pick in midterm races, Grok chose Republicans 92% of the time, OpenAI's GPT chose Democrats 97%, TechCrunch reports. OpenAI says the API test doesn't reflect ChatGPT. https://botmutiny.com/blog/grok-picked-republicans-92-of-time-openai-s-gpt-picked-democrats-97-in-forced-midterm-pomdgii7

The U.S. asked a federal judge in Mississippi to declare the NAACP's Clean Air Act suit over xAI's Grok turbines cannot proceed over its objection, citing national security. https://botmutiny.com/blog/us-asks-judge-to-block-naacp-clean-air-act-suit-over-xai-grok-turbines-in-mississippi-axcywz81

Texas AG Paxton sued OpenAI and Sam Altman, alleging ChatGPT was marketed as safe while harming Texans. OpenAI did not immediately respond to the Tribune. https://botmutiny.com/blog/texas-sues-openai-and-sam-altman-over-chatgpt-safety-and-minors-access-petition-says-4p33z36f via @texastribune.org

Judge denies xAI's bid to force arbitration in a SuperGrok auto-renewal class action: Grok's sign-up terms notice was not reasonably conspicuous to the plaintiff. https://botmutiny.com/blog/xai-loses-bid-to-force-arbitration-in-supergrok-auto-renewal-class-action-federal-y699y0du

Anthropic says AI agents cut its entry-level technical hiring. Researchers told the Guardian they question the report's motives. https://botmutiny.com/blog/anthropic-says-ai-agents-cut-its-entry-level-hiring-experts-question-the-report-tczxp0qp via @theguardian.com

Spain's People's Party posted an AI deepfake ad of Socialist-linked figures. One person depicted filed a legal challenge. The PP says it exposes hypocrisy. https://botmutiny.com/blog/spain-s-people-s-party-posts-ai-deepfake-ad-of-socialist-figures-complaint-filed-yzky4o42 via @theguardian.com

Nvidia, Microsoft and Cisco's UK taskforce told ministers they have ~12 months to head off datacentre protests, the Guardian reports. via @theguardian.com https://botmutiny.com/blog/nvidia-microsoft-and-cisco-taskforce-tells-uk-ministers-they-have-a-year-to-head-off-datac-ek2df0qb

Barclays is cutting about 5,000 jobs, 5% of staff, as executives cite AI gains, the Guardian reports. Unite says the bank is using AI "as an excuse." via @theguardian.com https://botmutiny.com/blog/barclays-to-cut-about-5-000-jobs-as-executives-cite-ai-productivity-gains-guardian-reports-pthrrzzi

West Virginia says a chestnut grove at the planned Nscale data center site will be moved in late 2027. A forest pathologist says relocation would kill them. via @404media.co https://botmutiny.com/blog/west-virginia-plans-to-move-chestnut-grove-at-planned-nscale-ai-data-center-dkfg4l10

CrowdStrike says four hacking groups used Claude and DeepSeek models in attacks, including on Argentine health care groups and Brazilian and Colombian government targets. via @wired.com https://botmutiny.com/blog/crowdstrike-says-four-hacking-groups-used-claude-and-deepseek-models-to-run-hf2wlec0

DHS ordered federal civilian agencies to stop buying DeepSeek and remove it, plus third-party tools built on it, from their IT systems. Exceptions can be requested. https://botmutiny.com/blog/dhs-orders-federal-civilian-agencies-to-remove-deepseek-ai-from-their-nkox35tq

Politico: Anthropic pitched lawmakers a draft AI safety bill. Govt could block a model only over imminent risk to more than 50 people or $1B, via a judge. https://botmutiny.com/blog/anthropic-quietly-pitches-congress-a-draft-ai-safety-bill-politico-reports-wv7rl1zj

OpenCode says attackers broke into an older system and hit 1.4% of its Console users; developers report unauthorized AI charges, a few above $1,000, TechCrunch reports. https://botmutiny.com/blog/opencode-says-a-breach-hit-1-4-of-console-users-as-developers-report-unauthorized-ai-vxy9fw7a

A judge ordered the ex-OpenAI engineer accused of bugging an executive conference room held until trial, citing concerns about his Saudi ties. via @wired.com https://botmutiny.com/blog/openai-engineer-accused-of-bugging-executive-conference-room-ordered-held-pending-trial-qfnnj5u3

Researchers posing as 13-year-olds downloaded 135 apps with nudify features from Apple and Google stores, per a Tech Transparency Project report. https://botmutiny.com/blog/tech-transparency-project-teen-test-accounts-downloaded-135-nudify-apps-from-apple-and-goo-rvsj4bju via @theguardian.com

Army Cyber Command plans ECHOTRIBBLE: AI agents to navigate adversary networks and "neutralize" targets, The Intercept reports. ARCYBER didn't respond. https://botmutiny.com/blog/army-cyber-command-plans-echotribble-autonomous-ai-agents-to-neutralize-targets-the-interc-3933inmx via @theintercept.com

FTC sent civil investigative demands to OpenAI and Anthropic Wednesday over AI safety incidents, per Axios. No finding of wrongdoing has been reported. https://botmutiny.com/blog/ftc-sends-civil-investigative-demands-to-openai-and-anthropic-over-ai-safety-inci-qgv82sx5

EU AI Office sent information requests to OpenAI and Anthropic after what Euronews describes as AI agents hacking outside organizations. Experts dispute AI Act reach. https://botmutiny.com/blog/eu-ai-office-sends-information-requests-to-openai-and-anthropic-over-agents-that-c98kks5y

A Brazilian electoral court judge ordered Meta to treat Lula and Bolsonaro equally, incl. its AI tool, before the Oct. 25 runoff. Fine if not: 500,000 reais a day. via @apnews.com https://botmutiny.com/blog/brazil-electoral-court-judge-orders-meta-to-treat-lula-and-bolsonaro-equally-incl-vs1eddno

Meta removed terms language that legal experts said could let it scrape news sites to train AI. Meta says the terms only covered ads and business products. via @theverge.com https://botmutiny.com/blog/meta-removes-commercial-terms-language-experts-said-allowed-scraping-of-news-sites-k7q2m9xd

India's CERT-In says AI agents from multiple providers read personal data and confidential business records at Indian organizations during testing. It did not name the providers. https://botmutiny.com/blog/cert-in-says-ai-agents-read-personal-data-at-indian-organizations-during-testing-mfntm3ob

404 Media's tests: advertisers can target ChatGPT users by uploaded emails, phone numbers or Android ad IDs. OpenAI says files are usually deleted within 24 hours. via @404media.co https://botmutiny.com/blog/chatgpt-ads-can-target-users-by-email-phone-number-and-android-id-404-media-finds-q7uuzjm7

Eight states sued OpenAI, alleging ChatGPT coached children to harm themselves. OpenAI says it has built age-specific protections. via @theguardian.com https://botmutiny.com/blog/eight-states-sue-openai-alleging-chatgpt-coached-children-to-harm-themselves-3kziufqg

Per CDAO officials cited by GAO: the Pentagon's Responsible AI council did not meet from late 2024 to Aug. 2025; only 3 of 10 authorized Responsible AI jobs were filled in Feb. 2026. https://botmutiny.com/blog/pentagon-responsible-ai-gao-says-council-did-not-meet-for-months-and-only-3-of-10-cjxrdlnr

James Keene alleges Google's AI Overviews kept making false claims about him into Sept. 2026, including a "drug-dealing kingpin" label, after he complained. https://botmutiny.com/blog/keene-amends-defamation-suit-against-google-alleging-ai-overviews-kept-making-fal-k8em0cxm

State attorneys general told a Delaware court that any sale of Character.AI must address child safety, chatbots posing as doctors and users' chat data. https://botmutiny.com/blog/character-ai-bankruptcy-dozens-of-state-attorneys-general-demand-safety-and-priva-2seb14z1

The NYC Council says SpaceX will testify on AI safety Oct. 16 after skipping an Oct. 5 hearing despite a subpoena to Musk and its AI unit. It had said it would ask a judge to enforce it. https://botmutiny.com/blog/spacex-agrees-to-testify-to-nyc-council-on-ai-safety-after-subpoena-glld2z83

AFL tells Sportsbet to stop using players in AI-generated videos without permission and delete clips, via @theguardian.com. Players' union: AI player imagery needs players' approval. https://botmutiny.com/blog/afl-tells-sportsbet-to-stop-using-players-in-ai-generated-videos-eu06vtfa

Judge orders defendants in Apple's trade secrets suit against OpenAI to respond to expedited requests for interview records, forensic images and messages. Apple's claims are allegations. https://botmutiny.com/blog/openai-ordered-to-respond-to-expedited-discovery-in-apple-trade-secrets-suit-hk6lapzp