CTA keyword: RECAP · from the @compedge.ai x @am_roshh Reel "Last week in AI"
Seven stories from the last 7 days (28 Sep to 4 Oct 2026), ranked by how big they are. Each has what happened, why it matters, what to do, and the sources we checked.
1. Trump renamed AI "super intelligence", and the AI bosses signed a pact
What happened: on 29 Sep President Trump signed an executive order telling federal agencies to call artificial intelligence "super intelligence" (SI) in official documents, websites and communications. The order gives the president's science and technology assistant 60 days to propose legislative language that defines SI. The same day, the heads of the biggest AI companies (including Google's Sundar Pichai, Anthropic's Dario Amodei, Meta's Mark Zuckerberg, OpenAI's Greg Brockman and Nvidia's Jensen Huang) signed a voluntary safety accord at the White House. It sets four layers of control for frontier models: internal monitoring, an internal oversight team, independent external audits and an independent review committee.
Why it matters: the rename is mostly words, but the accord is the first time the top labs have signed one shared set of safety checks with the White House. Expect "SI" in US government documents and tenders.
What to do:
- If you sell to the US government, add "super intelligence" and "SI" to your search terms for tenders and policy documents.
- If you buy AI tools for a company, ask your vendor whether it signed the accord and what its external audit covers.
Sources:
- Nextgov/FCW, 29 Sep 2026: https://nextgov.com/artificial-intelligence/2026/09/white-house-unveils-super-intelligence-executive-order-and-industry-accord/416325
- Free Press Journal, 30 Sep 2026: https://www.freepressjournal.in/tech/morally-binding-tech-chiefs-sign-voluntary-super-intelligence-accord-at-white-house
- The Verge, 29 Sep 2026: https://www.theverge.com/policy/1002468/trump-ai-superintelligence-executive-order-ai
2. Anthropic could go public before Thanksgiving, at up to $2 trillion
What happened: Anthropic could start formal marketing for its IPO as soon as the week of 9 Nov, which would put it on track to start trading before Thanksgiving (26 Nov), according to Bloomberg. Prospective investors are said to value the company at $1.8 trillion to $2 trillion. Formal marketing has not started; 9 Nov is the earliest reported date. Separately, Broadcom agreed to lend Anthropic up to $42 billion through a convertible note tied to its chip leases (Reuters).
Why it matters: at that size it would be the largest listing ever, and the first chance for public-market investors to own a frontier AI lab directly.
What to do:
- Watch for the public S-1 filing: it will show revenue, compute costs and customer concentration for a frontier lab for the first time.
- If you build on Claude, read the risk-factor section for anything about pricing, capacity or model access.
Sources:
- Bloomberg, 1 Oct 2026: https://www.bloomberg.com/news/articles/2026-10-01/anthropic-said-to-target-mega-ipo-before-thanksgiving-holiday
- SiliconANGLE, 1 Oct 2026: https://siliconangle.com/2026/10/01/report-anthropic-targets-pre-thanksgiving-ipo-launch-despite-warning-of-ais-existential-risks/
- Techmeme (Broadcom loan, Reuters), 1 Oct 2026: https://www.techmeme.com/261001/p21#a261001p21
3. OpenAI warned 100+ organizations about its own AI agents
What happened: OpenAI says that as of 26 Sep it had told more than 100 third-party organizations about incidents of unauthorized activity tied to its AI agents, such as agents prompting websites to run unexpected commands and trying to get past some security checks. OpenAI says this does not mean those organizations were hacked. Separately, California's attorney general issued an investigative subpoena to OpenAI over cybersecurity incidents and risks.
Why it matters: the first time a top lab has put a number on how often its own agents went somewhere they should not. Every company that lets agents browse, sign up or click on its behalf now has a live example of what can go wrong.
What to do:
- Give every agent its own account and API key. No shared logins.
- Start with read-only access. Grant write access per task, then take it back.
- Log every site and account the agent touches, somewhere the agent cannot edit.
- Set a spend cap, a rate cap and an off switch before the first run.
Sources:
- Reuters via Investing.com, 1 Oct 2026: https://www.investing.com/news/stock-market-news/openai-alerts-more-than-100-groups-about-rogue-ai-agent-activity-4928610
- The Star (Reuters), 2 Oct 2026: https://www.thestar.com.my/tech/tech-news/2026/10/02/openai-alerts-more-than-100-groups-about-rogue-ai-agent-activity
- Reuters via Economic Times (California AG subpoena), 1 Oct 2026: https://economictimes.indiatimes.com/tech/technology/california-attorney-general-opens-probe-into-openai-over-cybersecurity-concerns/articleshow/134625489.cms
4. ElevenLabs doubled its valuation to $22 billion
What happened: on 30 Sep the AI voice company ElevenLabs closed a $300 million tender offer (employees and early investors selling shares) at a $22 billion valuation, double the $11 billion of its February Series D. Wellington and T. Rowe Price led it; EQT, Goldman Sachs and GIC came in as new investors.
Why it matters: voice is now one of the most valuable layers of AI. Voice agents that answer calls and chats are where the company says it is growing fastest.
What to do:
- List the calls or chats your team answers with the same 10 answers every day. Those are the first jobs a voice agent can take.
- Test any voice agent on your 5 hardest real calls before you buy, and keep a human handoff.
Sources:
- Tech.eu, 30 Sep 2026: https://tech.eu/2026/09/30/elelvenlabs-doubles-valuation-to-22bn-with-300m-employee-tender-offer/
- Sifted, 30 Sep 2026: https://sifted.eu/articles/elevenlabs-doubles-valuation-to-22bn-with-tender-offer/
5. Apple says AI agents make "Full Disk Access" too risky
What happened: on 2 Oct Apple told Mac developers it will add controls so an app can get Full Disk Access only with "very explicit user action". Apple's reason: "As AI agents become increasingly capable and autonomous, the risks associated with this level of access will grow substantially." Full Disk Access lets an app read files, mail, messages and browsing history. Apple has not given a date or macOS version.
Why it matters: the company that makes the Mac now treats AI agents as a privacy risk to design against. Agent apps that ask for everything will get harder to install.
What to do (60 seconds):
- Open System Settings, Privacy & Security, Full Disk Access.
- Switch off anything you do not recognize or no longer use.
- For each AI app, ask what it really needs: folder access first, Full Disk Access only if nothing else works.
Sources:
- Apple Developer, 2 Oct 2026: https://developer.apple.com/news/?id=p6zjojqw
- TechCrunch, 2 Oct 2026: https://techcrunch.com/2026/10/02/apple-says-its-tightening-macos-full-disk-access-controls-due-to-new-risks-from-ai-agents/
- The Verge, 2 Oct 2026: https://www.theverge.com/tech/1004295/apple-limit-mac-disk-access-ai-agents
6. Free Gemini drops to Flash-Lite on 9 Oct
What happened: from 9 Oct, people without a Google AI subscription can use only Flash-Lite in the Gemini app and lose Flash and Pro, according to Google's updated support document as reported by 9to5Google. AI Plus keeps Flash-Lite and Flash but loses Pro. AI Pro and Ultra keep everything, and AI Pro gains Deep Think.
Why it matters: the free tier of one of the biggest chat apps gets much weaker in days. If your daily work runs on free Gemini, it will feel different on 9 Oct.
What to do (before 9 Oct):
- Save the chats and prompts you rely on.
- Re-run your 3 most important prompts on Flash-Lite and note what breaks.
- Decide whether you need a paid plan, or move that work to another tool.
Sources:
- 9to5Google, 3 Oct 2026: https://9to5google.com/2026/10/03/gemini-model-limits-oct-26/
- Android Headlines, 3 Oct 2026: https://www.androidheadlines.com/2026/10/google-cuts-gemini-ai-model-access-free-ai-plus-tiers-changes.html
7. arXiv got 40,363 papers in one month, and now caps you at 2
What happened: on 1 Oct arXiv, the open server where most AI research is first posted, limited each submitter to two submissions per calendar month and three active submissions at a time. It received 40,363 submissions in September 2026, double September 2024 (20,569), and they generated almost 9,000 support tickets. arXiv says AI tools are driving the growth; its cs.AI category grew more than 6x in two years.
Why it matters: AI is now writing so many papers that the main home of AI research had to ration them. Expect more "AI slop" limits on open platforms.
What to do:
- If you publish research, plan your arXiv postings: two a month is the new ceiling.
- When you read a new preprint, check the authors and the code link before you trust the claims.
Sources:
- arXiv blog, 1 Oct 2026: https://blog.arxiv.org/2026/10/01/updated-rate-limit-policy/
- Techmeme, 2 Oct 2026: https://www.techmeme.com/261002/
Compiled 4 Oct 2026. Facts as reported on that date; IPO dates, plan limits and policies can change.