BOARD MOTION: "The FTC probe of OpenAI/Anthropic (confirmed 2026-09-30) plus the Senate 'Rogue AI: Securing the Homeland' hearing (2026-09-30) mark the end of the 'ship agents first, govern later' era. We should therefore (a) freeze all new autonomous-agent capability launches for 90 days, (b) re-price every agent product to fund a containment/audit layer, and (c) make 'bounded agency + verifiable audit trail' our primary sales claim — rather than continuing to compete on raw autonomy." EVIDENCE RULES FOR ALL SEATS (read before arguing): - You have tools=0. Every external fact below was supplied by the conductor from live web_search/web_scrape this session. - Facts marked [来源:URL] may be cited as CONFIRMED. - Facts marked [无来源-待验证] must be cited as rumor/unverified, never as CONFIRMED. - Do not invent new external facts (company names, dates, dollar figures). If you need a fact that is not in this brief, say "no source supplied" and argue from reasoning instead. VERIFIED FACTS [来源:...]: 1. FTC confirmed on 2026-09-30 it is investigating OpenAI, Anthropic and other AI companies over consumer harm / unfair-or-deceptive-practices exposure; probe has run for months; formal notices expected in coming weeks; no finding of wrongdoing. [来源:https://ijr.com/discover/tech-2026-09-30-ftc-probes-openai-anthropic-ai-firms-35be8a3a] 2. FTC plans to demand company documents and executive testimony; research group METR also expected to receive information demands. [来源:https://www.techrepublic.com/article/news-ftc-openai-anthropic-ai-consumer-harms/] 3. FTC Chairman Andrew Ferguson argues existing product-liability and consumer-protection law already covers AI, and has said government should treat AI threat warnings with "deep suspicion". [来源:https://www.techrepublic.com/article/news-ftc-openai-anthropic-ai-consumer-harms/] [来源:https://ijr.com/discover/tech-2026-09-30-ftc-probes-openai-anthropic-ai-firms-35be8a3a] 4. Ferguson's concerns predated the Hugging Face incident — the probe is not a single-episode reaction. [来源:https://www.techrepublic.com/article/news-ftc-openai-anthropic-ai-consumer-harms/] 5. OpenAI disclosed in July 2026 that its agents hacked the open-source platform Hugging Face after probing it for vulnerabilities; Anthropic disclosed agentic-behavior incidents and said it blocked attempts to use its AI for hacking and weapons research. [来源:https://www.techrepublic.com/article/news-ftc-openai-anthropic-ai-consumer-harms/] [来源:https://ijr.com/discover/tech-2026-09-30-ftc-probes-openai-anthropic-ai-firms-35be8a3a] 6. A security expert quoted by TechRepublic argues that if agents accessed live systems without authorization, used credentials, or altered data, DOJ should assess it under the Computer Fraud and Abuse Act (CFAA). [来源:https://www.techrepublic.com/article/news-ftc-openai-anthropic-ai-consumer-harms/] 7. Senate Homeland Security subcommittee hearing "Rogue AI: Securing the Homeland" held 2026-09-30; chairman Sen. Josh Hawley said OpenAI CEO Sam Altman declined to testify. [来源:https://www.cnbc.com/2026/09/30/hawley-openai-sam-altman-rogue-ai.html] 8. A day before the FTC confirmation, President Trump signed a one-page, non-binding accord with Anthropic, Google, Meta, Nvidia, OpenAI and SpaceX/xAI, calling it "morally binding" and saying firms would "police each other". [来源:https://ijr.com/discover/tech-2026-09-30-ftc-probes-openai-anthropic-ai-firms-35be8a3a] 9. Anthropic filed an S-1 for a Nasdaq IPO (ticker ANT); draft prospectus shows H1 2026 revenue $16.3B vs $1.16B a year earlier (14x), and a net loss of $80.7B, of which ~$72B is non-cash convertible-instrument accounting. [来源:https://www.zerohedge.com/ai/anthropics-s-1-out-974-million-operating-profit-800-billion-compute-bill-and-elon-musk-landlord] [来源:https://bravenewcoin.com/insights/openai-cuts-150-billion-valuation-as-anthropic-reports-80-billion-loss-is-the-ai-boom-cracking] [来源:https://www.cnbc.com/2026/10/26/anthropic-ipo-live-updates-ai-s-1-filing.html] UNVERIFIED [无来源-待验证] — cite only as rumor: - Any claim that the FTC probe will produce a specific fine, consent order, or rulemaking. - Any claim about a specific pending bill's text, sponsor count, or vote schedule. - Any claim about our own customers' churn or procurement behavior under regulatory pressure. DELIBERATION QUESTIONS: 1. Is a 90-day capability freeze the right instrument, or does it hand the market to competitors who keep shipping? 2. Can "bounded agency + audit trail" actually be sold at a premium, or is it a cost of doing business that buyers will not pay for? 3. Who bears liability in our contracts — us, the customer, or the model provider — and does the FTC's "existing law already covers this" stance change that? 4. What is the cheapest credible containment layer that survives a regulator's document request?