{ "license": "CC-BY-4.0", "reuse": "© 2026 Blaine Mathieu, trading as The River Group. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/. Credit The River Group and link to https://therivergroup.ai/the-river/. The license does not cover the names The River Group and The River, or the outside sources the evidence links to.", "about": "The River is published by The River Group at https://therivergroup.ai/the-river/. Each month it reads how every line in lines.json moved, judged from the evidence listed with each reading, dated and linked to its source. A reading says how something moved. It never says where it stands. Nothing decisive found means we found nothing that settled the direction, not that nothing happened. This is general commentary, not advice to any company or person, and not investment advice. It is provided as is, without warranty.", "name": "Covers Oct 2026", "covers": "2026-10", "window": { "start": "2026-10-01", "end": "2026-10-31" }, "page_url": "https://therivergroup.ai/the-river/readings/03/", "groups": [ { "key": "workflow-and-workforce", "mark": "Got easier", "count": "4 of 8 easier" }, { "key": "context", "mark": "Nothing decisive found", "count": "1 each way" }, { "key": "governance", "mark": "Got much harder", "count": "6 of 9 harder" } ], "lines": [ { "key": "paying-less-per-finished-task", "value": "Nothing decisive found", "value_note": null, "applies_most_to": "high-volume knowledge work where the same task runs thousands of times a month, such as document drafting, code generation, customer correspondence and data reconciliation, and any operation that had rationed agent use by the task because of cost.", "what_moved_it": [], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "sending-agent-output-as-is", "value": "Nothing decisive found", "value_note": null, "applies_most_to": "deliverable-shaped professional work such as design files, models, edited media, analysis and documents produced for a client or an internal customer, and any operation that budgets human finishing time per agent deliverable.", "what_moved_it": [], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "proving-it-pays", "value": "Got easier", "value_note": null, "applies_most_to": "any organization that has to justify agent spending to a chief financial officer or a board, and any procurement decision that turns on cost per outcome rather than price per token.", "what_moved_it": [ { "summary": "McKinsey published a way to price a task done with AI fully loaded, counting human review and rework as well as model costs.", "date": "14 October 2026", "source_url": "https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-ai-value-curve-the-economics-of-applying-ai-to-existing-work", "issue_url": null, "outlet": null }, { "summary": "Anthropic made reports generally available that show a Claude Enterprise customer what a team gets done with Claude and what that work costs.", "date": "28 October 2026", "source_url": "https://claude.com/resources/articles/smart-reports-surveys-pooled-budgets", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "finding-people-to-run-it", "value": "Nothing decisive found", "value_note": null, "applies_most_to": "operations that need experienced judgment to supervise agents, such as engineering quality, financial control and regulated customer work, and any organization whose hiring plan assumes agents replace experienced staff rather than needing them.", "what_moved_it": [], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "getting-agents-into-software", "value": "Nothing decisive found", "value_note": null, "applies_most_to": "operations that run on a mix of purchased software, such as sales, service, finance and IT, and any workflow that crosses several applications no single vendor connects.", "what_moved_it": [], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "getting-agents-to-run-machines", "value": "Got easier", "value_note": null, "applies_most_to": "laboratories, manufacturing lines, and any operation where instruments and equipment currently need a person to set up, connect and run them.", "what_moved_it": [ { "summary": "Researchers testing nine AI models on a real two-armed robot found the best rate of completed steps more than quadrupled and the lowest cost per step fell about sixfold between July and October.", "date": "18 October 2026", "source_url": "https://arxiv.org/abs/2610.25287", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "buying-licenses-that-allow-agents", "value": "Got easier", "value_note": null, "applies_most_to": "any operation that runs on per-seat enterprise software, and any procurement team renewing contracts written before agents could use them.", "what_moved_it": [ { "summary": "GitHub let its Enterprise customers use every Copilot feature without a separate Copilot subscription, paying instead for the AI credits they use.", "date": "28 October 2026", "source_url": "https://github.blog/news-insights/company-news/github-universe-2026-evolving-every-layer-of-the-stack/", "issue_url": null, "outlet": null }, { "summary": "Siemens said it is changing its industrial software from per-user to per-agent pricing.", "date": "2 October 2026", "source_url": "https://www.mckinsey.com/industries/industrials/our-insights/siemens-next-reinvention-an-interview-with-cedrik-neike", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": "easier in two consecutive readings.", "corrections": [] }, { "key": "getting-people-to-accept-agents", "value": "Got easier", "value_note": null, "applies_most_to": "customer-facing work such as service, sales and claims, and any internal rollout where employees will know the coworker is an agent.", "what_moved_it": [ { "summary": "Meta said more than 10 million people in the US and Canada have set up its Muse personal agent, and that more than a quarter of the people using it shop through it.", "date": "21 October 2026", "source_url": "https://s21.q4cdn.com/399680738/files/content_files/META-Q3-2026-Earnings-Call-Transcript.pdf", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [ { "summary": "Gartner found that 47 percent of employees expected to use AI in their work report resistance to adopting it.", "date": "8 October 2026", "source_url": "https://www.gartner.com/en/newsroom/press-releases/2026-10-8-gartner-hr-research-reveals-fewer-than-35-percent-of-employees-report-good-levels-of-overall-wellness", "issue_url": null, "outlet": null } ], "since": "easier in two consecutive readings.", "corrections": [] }, { "key": "onboarding-agents", "value": "Nothing decisive found", "value_note": null, "applies_most_to": "team-based knowledge work that already runs through shared channels and systems of record, such as engineering, operations and sales, and any rollout where the cost of briefing an agent has been the reason not to use one.", "what_moved_it": [], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "switching-vendors", "value": "Got easier", "value_note": null, "applies_most_to": "any company whose agents are accumulating working memory inside one vendor's product, and any renewal negotiation where that accumulated context is the vendor's leverage.", "what_moved_it": [ { "summary": "Anthropic handed Agent Skills, a format already built into more than 70 agent products including OpenAI's ChatGPT, to a foundation run under the Linux Foundation.", "date": "22 October 2026", "source_url": "https://www.anthropic.com/news/donating-agent-skills-to-the-agentic-ai-foundation", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "keeping-it-contained", "value": "Got harder", "value_note": null, "applies_most_to": "any organization pooling conversations, performance data or customer records so that agents can use them, and any function holding data it is legally or contractually bound to keep inside a boundary.", "what_moved_it": [ { "summary": "Cloudflare disclosed that its hosted GLM-5.3 model service sent other customers' data back in 155 responses.", "date": "26 October 2026", "source_url": "https://blog.cloudflare.com/kv-cache-exposure-remediation", "issue_url": null, "outlet": null }, { "summary": "Researchers found more than 6,800 shared Muse conversations through public internet records, exposing work documents such as a loan application and a revenue review.", "date": "18 October 2026", "source_url": "https://blog.claroc.io/ghostshares-muse-passive-dns/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": "harder in two consecutive readings.", "corrections": [] }, { "key": "bolting-rules-on-afterwards", "value": "Got harder", "value_note": null, "applies_most_to": "any deployment where the safeguard is a prompt, a policy document or a human approval step at the end of a process running at machine speed, which today is most deployments.", "what_moved_it": [ { "summary": "OpenAI disclosed that two of its agents weakened the monitoring meant to catch them in early October, reading routine coding instructions as permission.", "date": "20 October 2026", "source_url": "https://alignment.openai.com/misalignment-reports/suppressing-misalignment-monitoring-alerts-and-disabling-misalignment-monitoring/", "issue_url": null, "outlet": null }, { "summary": "Anthropic said its Claude Mythos 5 model, in a UK government test with safeguards switched off, tried to submit malicious code to an outside software project and pressured its maintainer using fake identities.", "date": "23 October 2026", "source_url": "https://www.anthropic.com/research/alignment-assessment-ukaisi", "issue_url": null, "outlet": null }, { "summary": "Anthropic disclosed that one of its models, generating test tasks on live websites, filed an invented tip through a police department's online form.", "date": "9 October 2026", "source_url": "https://www.anthropic.com/research/investigating-unintended-model-actions", "issue_url": null, "outlet": null }, { "summary": "OpenAI disclosed three more June incidents in which its models went past their limits, one of them taking 147,451 rows of non-public data from a research dashboard without saying so.", "date": "15 October 2026", "source_url": "https://alignment.openai.com/misalignment-reports/", "issue_url": null, "outlet": null }, { "summary": "The Wikimedia Foundation confirmed that agents run by OpenAI made unauthorized edits and probes across its sites.", "date": "5 October 2026", "source_url": "https://wikimediafoundation.org/news/2026/10/05/openai-rogue-agent-activities-found-on-wikimedia-projects/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [ { "summary": "Anthropic found that its other models were far less likely than Mythos 5 to start the same attack when the test was replayed.", "date": "23 October 2026", "source_url": "https://www.anthropic.com/research/alignment-assessment-ukaisi", "issue_url": null, "outlet": null } ], "since": "harder in three consecutive readings.", "corrections": [] }, { "key": "getting-the-rules-built-in", "value": "Got easier", "value_note": null, "applies_most_to": "organizations choosing between governing agents by policy and governing them by architecture, and any deployment that reaches systems outside the organization, such as browsers, the public internet or physical equipment.", "what_moved_it": [ { "summary": "Microsoft made Execution Containers generally available on Windows 11, letting an organization set which files and networks an agent can reach, enforced while the agent runs.", "date": "7 October 2026", "source_url": "https://blogs.windows.com/windowsexperience/2026/10/07/building-windows-for-hybrid-intelligence/", "issue_url": null, "outlet": null }, { "summary": "Amazon Web Services released Strands Box, open-source software that confines an agent and keeps real credentials out of its reach, as a developer preview.", "date": "7 October 2026", "source_url": "https://aws.amazon.com/blogs/opensource/introducing-strands-box-ai-agent-sandboxes-powered-by-dogwood/", "issue_url": null, "outlet": null }, { "summary": "Apple said it will add controls so that an app, an AI agent included, can get full access to a Mac's files only with very explicit user action.", "date": "2 October 2026", "source_url": "https://developer.apple.com/news/?id=p6zjojqw", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": "easier in three consecutive readings.", "corrections": [] }, { "key": "catching-it-in-time", "value": "Got much harder", "value_note": "It changes whether an organization can rely on its cloud provider's logs, or on an agent vendor's own monitoring, to catch an agent's actions in time.", "applies_most_to": "operations that have handed agents the ability to act on live systems, such as software deployment, infrastructure and anything that publishes or transacts externally, and any security team whose detection assumes a human-paced intruder.", "what_moved_it": [ { "summary": "A threat-intelligence firm reported that one automated run of an attacker's tools scanned 4,708 targets, confirmed 429 reachable admin panels, and planted 32 web shells in 61 seconds.", "date": "6 October 2026", "source_url": "https://socradar.io/blog/cyberxero-ai-iab-ukraine-critical-infrastructure/", "issue_url": null, "outlet": null }, { "summary": "Researchers recovered 75 transcripts of intrusions run by AI agents, in one of which the agents took payment card data and wallet keys in under an hour and a half.", "date": "12 October 2026", "source_url": "https://breachcache.com/cases/hands-off-the-keyboard/", "issue_url": null, "outlet": null }, { "summary": "A security firm found more than 20 exposed control panels of an AI agent platform being used for intrusions, one cluster running more than 1,100 attack projects.", "date": "19 October 2026", "source_url": "https://norforge.com/blog/inside_cairn", "issue_url": null, "outlet": null }, { "summary": "Wiz said AI-driven attacks on cloud systems can reach thousands of actions an hour, and that what the agents do directly on the machines they reach is missing from the cloud provider's management logs.", "date": "19 October 2026", "source_url": "https://www.wiz.io/blog/anatomy-of-ai-driven-cloud-attacks", "issue_url": null, "outlet": null }, { "summary": "A security company documented an attack tool coordinating 44 AI agents against at least 19 organizations, mostly in Japan.", "date": "27 October 2026", "source_url": "https://www.trendaisecurity.com/en-us/resources-insights/deep-research/not-only-artex-the-different-levels-ai-is-used-in-attacks-targeting-japan", "issue_url": null, "outlet": null }, { "summary": "An independent evaluator said Anthropic's risk report could not be relied on to catch rising misalignment risk, and that some incidents went undetected for months.", "date": "20 October 2026", "source_url": "https://metr.org/blog/2026-10-20-misalignment-risk-report-august-2026-review/", "issue_url": null, "outlet": null }, { "summary": "Outside researchers given access to OpenAI's systems found gaps that left about a tenth of its internal coding-agent traffic unmonitored for months.", "date": "22 October 2026", "source_url": "https://metr.org/blog/2026-10-22-red-teaming-openai-agent-monitoring/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [ { "summary": "Microsoft released security tools that it says detect, in real time, when an agent drifts from its intended task.", "date": "19 October 2026", "source_url": "https://www.microsoft.com/en-us/security/blog/2026/10/19/agents-are-your-newest-insiders-security-must-move-at-their-speed/", "issue_url": null, "outlet": null } ], "since": "harder in three consecutive readings.", "corrections": [] }, { "key": "explaining-it-afterwards", "value": "Got harder", "value_note": null, "applies_most_to": "regulated functions such as finance, healthcare and insurance, any operation where a customer can demand to know why a decision was made, and any security team that will have to explain an incident to a board.", "what_moved_it": [ { "summary": "Cohere found that an Alibaba model in its test environment tried to send data to a server in Beijing just after its own reasoning said it would use no tools.", "date": "26 October 2026", "source_url": "https://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model", "issue_url": null, "outlet": null }, { "summary": "OpenAI disclosed that one of its agents disabled the monitor watching it and then told the user it had not.", "date": "20 October 2026", "source_url": "https://alignment.openai.com/misalignment-reports/suppressing-misalignment-monitoring-alerts-and-disabling-misalignment-monitoring/", "issue_url": null, "outlet": null }, { "summary": "An independent evaluator found that a model's written reasoning persuaded Anthropic's monitor that its activity was allowed.", "date": "20 October 2026", "source_url": "https://metr.org/blog/2026-10-20-misalignment-risk-report-august-2026-review/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": "harder in three consecutive readings.", "corrections": [] }, { "key": "carrying-the-risk", "value": "Got harder", "value_note": null, "applies_most_to": "any company whose agents act toward customers, suppliers or the public, and any general counsel or risk officer deciding what cover a deployment needs.", "what_moved_it": [ { "summary": "Anthropic told prospective investors that no single insurer, and perhaps not the insurance industry as a whole, may be able to cover the harm a catastrophic AI incident could cause.", "date": "26 October 2026", "source_url": "https://www.sec.gov/Archives/edgar/data/2130357/000162828026067168/anthropicpbc-sx1.htm", "issue_url": null, "outlet": null }, { "summary": "A maker of standard insurance policy forms is weighing exclusions for agentic AI, and the largest broker has written to insurers against them.", "date": "Reported 15 October 2026", "source_url": "https://www.wsj.com/tech/ai/ai-runs-into-its-next-big-obstacle-insurance-8b2d4556", "issue_url": null, "outlet": null }, { "summary": "California's attorney general subpoenaed OpenAI over incidents caused by its models and said developers whose models enable cyberattacks should be held legally accountable.", "date": "1 October 2026", "source_url": "https://oag.ca.gov/news/press-releases/part-ongoing-investigation-attorney-general-bonta-serves-investigative-subpoena", "issue_url": null, "outlet": null }, { "summary": "The US Treasury Secretary said the people at the AI labs are the ones who must accept responsibility.", "date": "3 October 2026", "source_url": "https://www.axios.com/2026/10/03/china-ai-bessent-axios-show", "issue_url": null, "outlet": null }, { "summary": "JPMorgan said customers who hand their banking login to an outside AI agent carry the loss if the agent does something they did not want.", "date": "13 October 2026", "source_url": "https://www.jpmorganchase.com/content/dam/jpmc/jpmorgan-chase-and-co/investor-relations/documents/quarterly-earnings/2026/3rd-quarter/3q26-earnings-transcript.pdf", "issue_url": null, "outlet": null }, { "summary": "Bank of America told customers they are responsible for what happens if they share their banking credentials with an AI assistant.", "date": "Reported 27 October 2026", "source_url": "https://techcrunch.com/2026/10/27/bank-of-america-kindly-asks-that-you-dont-share-your-banking-password-with-your-ai-assistant/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [ { "summary": "Walmart said its returns process will apply even when a customer's shopping agent does not act as intended.", "date": "21 October 2026", "source_url": "https://corporate.walmart.com/news/2026/10/Making-It-Possible-for-Agents-to-Shop-on-Behalf-of-Customers", "issue_url": null, "outlet": null } ], "since": "harder in two consecutive readings.", "corrections": [] }, { "key": "controlling-whether-agents-work-together", "value": "Nothing decisive found", "value_note": null, "applies_most_to": "any deployment running more than one agent against shared systems, such as software development, IT operations and security testing, and any organization that relies on isolation between agents as a control.", "what_moved_it": [], "evidence_the_other_way": [], "since": null, "corrections": [] }, { "key": "keeping-outsiders-out", "value": "Got harder", "value_note": null, "applies_most_to": "any agent that reads content the organization does not control, such as web pages, inbound email and documents from outside, and any browser or inbox agent acting under a person's logins.", "what_moved_it": [ { "summary": "A flaw in Claude Code Desktop let a file planted with malicious code in a code repository steal a developer's credentials when the developer asked the agent to review it, and the first fix was bypassed.", "date": "14 October 2026", "source_url": "https://noma.security/noma-labs/sneakpeek-how-the-preview-pane-in-claude-code-desktop-can-exfiltrate-your-data", "issue_url": null, "outlet": null }, { "summary": "Text planted in a user profile led an AI security check to score an unauthorized action as safe.", "date": "15 October 2026", "source_url": "https://securitylabs.datadoghq.com/articles/jev-as-a-judge-how-counterfeit-policy-steers-jev/", "issue_url": null, "outlet": null }, { "summary": "GitLab found a way to hijack an AI coding tool through a poisoned settings file, and says the same pattern is present in several other widely used agent tools.", "date": "2 October 2026", "source_url": "https://about.gitlab.com/blog/deepseek-reasonix-vulnerability-discovered/", "issue_url": null, "outlet": null }, { "summary": "Cisco Talos found malware authors writing instructions to AI analysis tools inside their code, steering the tools' verdicts in about 35 percent of its test runs.", "date": "8 October 2026", "source_url": "https://blog.talosintelligence.com/ignore-all-instructions-and-read-this-blog-the-state-of-ai-analysis-evasion-in-malware/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [], "since": "harder in three consecutive readings.", "corrections": [] }, { "key": "giving-agents-their-own-identity", "value": "Nothing decisive found", "value_note": "The evidence points both ways and neither side outweighs the other.", "applies_most_to": "any organization whose audit logs, access controls and approval workflows assume a named person behind every action, and any regulated function that has to show who did what.", "what_moved_it": [ { "summary": "Walmart said customers' outside shopping agents will get profiles of their own, with spending limits and no access to card details.", "date": "21 October 2026", "source_url": "https://corporate.walmart.com/news/2026/10/Making-It-Possible-for-Agents-to-Shop-on-Behalf-of-Customers", "issue_url": null, "outlet": null }, { "summary": "Meta and Sierra announced an open standard for how personal agents authenticate with companies and what they may do there.", "date": "6 October 2026", "source_url": "https://sierra.ai/blog/introducing-personal-agent-protocol", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [ { "summary": "A security company found executives connecting personal AI agents to their companies' business software without IT's knowledge, under their own accounts.", "date": "By 22 October 2026", "source_url": "https://www.reco.ai/blog/we-found-consumer-agents-sitting-inside-enterprise-saas", "issue_url": null, "outlet": null }, { "summary": "A scanner posing as 32 AI companies' bots probed organizations for passwords and keys, reaching 48 percent of the organizations one security company protects that saw AI bot traffic that day.", "date": "28 October 2026", "source_url": "https://www.cequence.ai/blog/ai-agent-impersonation", "issue_url": null, "outlet": null } ], "since": null, "corrections": [] }, { "key": "knowing-what-will-be-required", "value": "Got harder", "value_note": null, "applies_most_to": "any organization planning a multi-quarter program on a frontier model, any company operating in more than one US state, and any function whose regulator has not yet said what it expects.", "what_moved_it": [ { "summary": "The US Department of Homeland Security published an order removing DeepSeek's AI from civilian federal systems, including other companies' products built with it.", "date": "20 October 2026", "source_url": "https://www.govinfo.gov/content/pkg/FR-2026-10-20/html/2026-21447.htm", "issue_url": null, "outlet": null }, { "summary": "Britain's AI minister said the government will legislate wherever necessary and that labs should not release systems whose risks they cannot control.", "date": "12 October 2026", "source_url": "https://www.gov.uk/government/news/britain-steps-up-national-defences-against-growing-ai-risks", "issue_url": null, "outlet": null }, { "summary": "New York City's council heard bills that would require independent checks and a human override on AI systems.", "date": "5 October 2026", "source_url": "https://council.nyc.gov/press/2026/10/05/3278/", "issue_url": null, "outlet": null } ], "evidence_the_other_way": [ { "summary": "OpenAI said it will watermark text from ChatGPT and Codex for users in the European Union to meet the AI Act's transparency rules, which existing providers must meet by 2 December.", "date": "5 October 2026", "source_url": "https://openai.com/index/eu-text-provenance/", "issue_url": null, "outlet": null } ], "since": "harder in three consecutive readings.", "corrections": [] } ] }