Top 10 AI News This Week — October 26–November 1, 2026
Nov 2, 2026
The federal government stopped calling it artificial intelligence, and almost nobody noticed. An executive order signed in September now has agencies writing "Super Intelligence" and "SI" instead — which is why the Department of Education's new accreditation rule looked AI-free last week when it was not, and why a Justice Department indictment this week says fraudsters used "super intelligence" to do a student's homework. Meanwhile four AI companies agreed on a single opt-out for training on your website, and set the default to yes. Here's what community college educators should know.
EdTech & Campus AI
The federal government renamed AI, and it broke our own search — a correction
Executive Order 14434, "Inaugurating the Era of Super Intelligence," was signed September 29, 2026 and published October 2 at 91 FR 63129. Section 2(a) directs agencies, "to the maximum extent permitted by law," to use "Super Intelligence" and "SI" in place of "Artificial Intelligence" and "AI" in correspondence, public communications, websites, reports and policy documents. Section 2(b) exempts previously issued regulations, Presidential actions, contracts, grants and historical documents. It binds executive-branch agencies in nonstatutory documents only. It does not reach colleges, grantees or contractors.
The correction we owe readers: this digest has twice recorded ai.gov serving "SI" where "AI" belonged as a site defect. It was never a defect. It was compliance, and treating it as a typo cost us a story — last week we reported that the Department of Education's accreditation rule contained no AI provision. It does. It appears under the substituted term, and a keyword search for "artificial intelligence" missed it.
Three in-window confirmations that the change is live and spreading: ED's accreditation rule carries a footnote citing EO 14434 for the substitution; a Treasury proposed rule at 91 FR 67396 (October 26) does the same; and the FTC's October 28 pricing inquiry uses a third spelling, "superintelligence algorithms."
For community colleges the consequence is methodological and it is severe. Any search of federal documents keyed on "AI" now returns false negatives — grant notices, rules and guidance that govern campus technology increasingly say "SI." Local AI policies, procurement templates and syllabus language that define "AI" as a term will no longer match the federal vocabulary they were written to track. Watch for the Section 3(b) legislative language, due to the President around November 28, which is the thing that could push the term past nonstatutory documents and into grants. Read more →
A provost resigned over AI use, and the evidence was his institutional chat logs — not a detector
Dartmouth provost Santiago Schnell resigned Wednesday, October 28, after an independent committee of faculty and alumni found he "used AI to meaningfully alter his writing" and "fell short" of Dartmouth's academic integrity standards. The committee's eight-page report reached President Sian Leah Beilock on October 27; the resignation came the next day, less than two weeks before a scheduled faculty no-confidence vote.
The detail that matters is what the committee declined to use. The story began with detectors — a Semafor investigation reported that Pangram flagged Schnell's August 9 op-ed as at least 80% AI-generated, and the student paper later used the same tool to report nine of his 2026 articles as 96% "AI-written." The committee relied on none of it. It reviewed publishers' AI-disclosure policies and Schnell's institutional Dartmouth Claude logs, which showed he used Claude to "substantially edit" a research manuscript. For one forthcoming article he asked Claude whether there were "many signatures of AI writing," revised accordingly, and asked the editor to pause publication. The committee also found he did not fully cooperate, declining to supply logs from personal accounts. Schnell said AI did not substitute for his "research, intellectual judgment or authorship."
If you teach writing or sit on an academic integrity committee, this is the evidentiary standard argument arriving at the top of an institution: enterprise logs the college already holds proved decisive where detector percentages did not. It pairs with a thread we owed a verification pass — Cambridge refused to sign Turnitin's new licence in July 2026 over terms that would have let student work develop AI tools, and Southampton is phasing Turnitin out after 2026–27 (primary dated October 2, 2026, correcting the October 15 date we previously carried). Read more →
830,840 flagged student-aid IDs, five indictments — and an AI claim the charging documents don't make
The Justice Department announced Operation Ghostbusters on October 29 (release 26-1278): five defendants in three districts in a "ghost student" financial-aid scheme tied to a Kenyan network. District of New Jersey — three defendants, intended loss $1.5 million. Eastern District of Virginia — one sealed defendant, over 700 refund deposits totaling roughly $1,439,548. District of Nebraska — Francis Mogambi, ten counts of theft of government money and aggravated identity theft. DOJ puts the scale at "well over $200 million" and says investigators identified 830,840 potentially fraudulent FSA IDs relating to nearly $130 million. The FBI has so far identified $32 million in losses; over $750,000 in cryptocurrency was seized in what DOJ calls "the first international seizure of its kind in Kenya."
Read the AI framing carefully. DOJ's release never uses the words "artificial intelligence." It says twice, without explanation, that fraudsters use "super intelligence" — which, per the story above, is now the federal term. Press coverage diverged: ABC7 Los Angeles attributed an AI claim to "federal officials," while ABC News national omitted AI entirely. The method DOJ actually describes is identity theft — stolen identities of real Americans used to file FAFSAs and enroll victims — not synthetic students.
For California community colleges this is the enrollment-integrity number to watch rather than the AI angle. ABC7 reports the state's 116 colleges among the hardest hit, with roughly a third of 2024 applications fraudulent. DOJ recommended ED disable those 830,840 FSA IDs; ED has issued no rule, Dear Colleague Letter or guidance in response, which makes the federal follow-on the thing to expect. A separate in-window item is already operational: Dear Colleague Letter GEN-26-06 (October 26) sets 2027–28 verification items, with identity verifiable by in-person check or video call. Read more →
The California drought breaks: 40 paid AI internships across the system, and a Title III AI certificate down the road
After four weeks of searching for any new California community college AI program and finding none, two items landed. Axim Collaborative awarded two $750,000 grants — one to the California Community Colleges, one to the UNC System — reported October 28–29. On the California side, 40 students will receive a paid internship on an infrastructure project at one of the system's 116 campuses, with Amazon Web Services and Microsoft Elevate as industry partners. Cameron Sublett of the Foundation for California Community Colleges called the model "applied co-intelligence" — technical training plus durable skills plus AI use in a field — and said plainly that institutions are "kind of tired of building these really wonderful but boutique programs." Axim's Philipp Schmidt framed the goal as "a national standard for AI technician pathways."
Separately, and four days outside our window, Moorpark College announced a five-year, $3 million Title III Strengthening Institutions Program grant on October 22: a new interdisciplinary Applied Artificial Intelligence certificate students pair with any major, AI professional development for at least 50 faculty across 10 or more disciplines, an Applied AI Advisory Board for internships, and a target of 100 students in the certificate by year five. We run it late because it is the first California community college with an AI-specific SIP award, in the same cluster as the three we covered last week.
For community colleges: there is no published application route for the Axim internships — 40 seats across 116 colleges will be allocated by the system, so the action is an inquiry to the Chancellor's Office, not a form. Moorpark is the more useful document. A pair-with-any-major certificate plus 50-faculty development is the closest available template for a Fullerton Title III proposal. Read more →
Major AI Companies
Four AI companies agreed on one way to opt out of training — and the default is yes
On October 30, Anthropic, Google, Microsoft and OpenAI all backed the same robots.txt signal for AI training: the user-agent token __usage_ai-training. Anthropic's Kevin Kelley filed the underlying IETF Internet-Draft the same day; it reuses RFC 9309's path matching and precedence and adds no new directives. Note what it is not: the Datatracker page states the draft is not endorsed by the IETF and has no formal standing, it expires May 3, 2027, and the draft says consumers "MAY" refrain from retrieving content — a preference, not an access control.
The load-bearing sentence is Microsoft's, verbatim: "Training and grounding are set independently, and both default to Allow when no preference is published." A college that does nothing has consented. A wildcard User-Agent: * group does not express a training preference — the token must be named explicitly. And the shared token covers training only; answer-time grounding is separate, with Microsoft using page-level meta tags instead.
Dates differ sharply by company. Microsoft honors the preferences from February 1, 2027, with temporary grounding preferences opening in Bing Webmaster Tools December 15, 2026 and a March 1, 2027 deadline to publish before NoArchive and NoCache stop being read as signals. Google says "in 2027," its crawler docs say "early 2027." OpenAI says "in the coming months." Anthropic says it has already started.
This is the week's one item that needs a decision from a specific person on your campus. Whoever controls the district web server has until roughly March 1, 2027 to decide whether the college catalog, library guides, OER repositories and faculty pages are opted out — and silence is consent under all four defaults. Keep existing GPTBot and Google-Extended rules in place; the new token does not replace them, and it does not remove anything already collected. It is also a clean teaching case for a web or CIS course on how robots.txt actually works. Read more →
Anthropic filed to go public, and education is not in the prospectus
Anthropic filed its S-1 with the SEC on October 26 (accession 0001628280-26-067168, "Anthropic, PBC"), applying to list Class A shares on the Nasdaq Global Select Market under the symbol "ANT." The figures, read off the filing rather than press summaries: total revenue of $41 million in 2023, $386 million in 2024 and $4.585 billion in 2025, then $16.286 billion for the six months ended June 30, 2026 against $1.163 billion in the same period a year earlier. Net loss over those same six months: $80.748 billion, against $41.965 billion for all of 2025.
The public-benefit structure is disclosed as a risk. The prospectus states the PBC form "may effectively eliminate the ability of holders of our Class A common stock to influence important decisions," and that the company must "balance the pecuniary interests of our stockholders" against its stated purpose — "the responsible development and maintenance of advanced AI for the long-term benefit of humanity." The Long-Term Benefit Trust is described as five "financially disinterested voting trustees" with no day-to-day authority. On pricing, the filing commits to nothing: "We have also changed, and may in the future change, usage limits, rate limits, or model and product capabilities."
For community colleges the notable thing is an absence. Across the prospectus summary, the offering, the selected financials and the opening risk factors, there is no mention of students, schools, colleges, universities, FERPA, teachers or Claude for Education. "Education" appears once, as a forecast benefit of the technology. We did not read the full filing, so we state it as true of the summary and risk factors rather than of the whole document — but a company whose education tier many colleges are piloting did not consider that segment material enough to name to investors. Two companion items: Anthropic's OneGov 40% discount now extends to state, local, tribal and territorial governments through GSA's Multiple Award Schedule (October 27), and its 2026 Usage Policy takes effect November 12, requiring human review and disclosure for admissions ranking, academic misconduct decisions and credentialing exams. Read more →
What landed in education tenants this week, and what admins can and cannot switch off
Two vendors moved on education tiers in opposite directions. Microsoft's "Education for Cowork" — published October 20, and the gap we failed to close last week — is now documented on Microsoft Learn. Eligibility, verbatim: "faculty and staff users in the education industry can access Cowork. Students aren't eligible for access at this time." It requires an active Microsoft 365 Copilot license plus usage-based billing enabled by an admin before anyone can open it. Billing is consumption-based, not per seat: policies must be "at least 2,000 credits per user per month," and when users hit a limit "they lose access to agents and services for the rest of the month." No dollar-per-credit rate is published anywhere, so a district cannot currently build a budget line for it. There is no free tier and no education grant. Microsoft does disclose that Cowork "can use Anthropic Claude models as a subprocessor," and that admins can turn individual models off.
Google went the other way. Its new immersive explorations feature (October 28) reaches the Gemini student hub on "Education Fundamentals, Standard, and Plus" — including the no-cost edition most colleges run — and the post states twice: "There is no admin control for this feature" and "There is no end user setting for this feature." The only lever is turning the entire Gemini app off for the domain. The same posture appears on Gemini editing in Slides (October 30), which is available via a paid "Google AI Pro for Education" add-on rather than the base editions.
One Google item is worth chasing on its own merits: real-time speech translation in Meet now covers 63 languages with automatic language detection, translating into each listener's preferred language and handling a speaker who switches mid-sentence. The Scheduled Release rollout starts today, November 2. It is on by default for eligible editions but off at the start of each meeting — participants switch it on. Honest cost note: it requires the paid Google AI Pro for Education add-on, so it is not free on a bare Fundamentals license. For ESL, DSPS and multilingual outreach it is the most useful thing either vendor shipped this month. Read more →
Free & Student Opportunities
Free credentials, priced honestly — nothing launched, two long-open questions closed, and one offer quietly expired
No free credential launched in window. That is the finding, not a gap in the sweep. But three things changed.
The Gemini faculty exam exists. We have been asking for weeks. Google lists three Gemini certifications — Educator (K-12), Faculty ("designed for university faculty") and University Students — and states that "registration for all Gemini certification exams is free of charge." Valid three years, two hours, launch within 14 days of registering. Honest caveat: Google's words are "university faculty," and nothing on any readable Google page includes or excludes community college faculty. Nothing blocks a CC instructor from registering. The University Student credential remains the most CC-relevant free, resume-bearing credential we have found.
OpenAI's four free months of ChatGPT Plus for students expired October 31 and was not extended. The help article still reads "Claim the offer by October 31, 2026" with no closure notice. Treat it as closed. While you are there, a correction worth printing: OpenAI Academy course certificates "are not OpenAI Certifications" and "do not represent a formal OpenAI credential" — the actual certification is invite-only for Enterprise and Edu workspaces, so a student cannot self-enroll.
Also new to this beat: Anthropic now offers "Teaching AI Fluency" free on Skilljar for "academic faculty, instructional designers, and others," with an optional assessment and certificate of completion (undated, so we cannot date the launch). And an addition to our own record: AWS microcredentials are genuinely free and grant a Credly badge from a graded hands-on lab — distinct from the two free Skill Builder courses we covered on October 20, which grant nothing. Microsoft Applied Skills works the same way; Microsoft Learn badges and trophies are, in Microsoft's own words, "achievements, not credentials." Google AI Essentials is still $49/month after a seven-day trial. Claude for Teachers remains free only for verified US K-12 educators — CC faculty do not qualify, so don't chase it.
Finally, a template rather than an opportunity: San Diego CCD's Academic Applications of AI micro-credential is free, self-paced, two to four hours, and grants a digital badge — in student and faculty/staff versions, with participation data refreshed October 30. Eligibility is SDCCD only. A California district already shipped the thing Fullerton keeps discussing. Read more →
Two NSF deadlines in two days, a federal docket closing December 3, and this month's scholarship check comes up empty
Closing Wednesday, November 4, "by 5 p.m. submitting organization's local time." Two NSF solicitations, both naming community colleges verbatim as eligible: NSF 26-513 AI Infrastructure Hubs ($40–100M anticipated, ~10 awards, 5 years at $4–12M, and "only one award per state or multi-state region") and NSF 26-512 AI Datasets ($60–100M, 25–50 awards, with a $200,000 planning tier that is the only realistic single-college entry point). Both eligibility clauses read "Two- and four-year IHEs (including community colleges)." At two days' notice these are next-cycle targets — November 3, 2027 — and 26-513 names planning proposals as the stated on-ramp.
The genuinely actionable deadline is a comment docket. NCES published a Request for Information on October 29 (91 FR 68118) on redesigning its entire portfolio of education data collections, early childhood through adult. Question 4, verbatim: "How might data collection and dissemination be optimized for future Super Intelligence or SI." Comments are due December 3, 2026, there is no eligibility bar, and everything touching IPEDS and postsecondary reporting is in scope. An institutional research office can file. Cite 91 FR 68118 rather than the docket number — the notice as published carries an unfilled placeholder, "ED-2027-XXX-XXXX."
Two named leads resolved. The Gates Foundation AI Transformation Learning Grants cohort launches today — eight institutions at $150,000–$200,000 each, with "broad-access higher education institutions—including community colleges" named verbatim as eligible — but recipients were notified privately October 2 and no list has been published. And HRCC Round Two is real: the California Workforce Development Board released it October 28 at a confirmed $19 million, 13 awards, closing December 3 at 5:00 p.m. PST — but "Round Two funding is available to existing HRCC regions only," it never mentions colleges, and it has no AI dimension at all. "WAF 15" does not exist; the CWDB program that does name community colleges is the Accelerator series, and Accelerator 14 closed August 27.
This month's AI scholarship check, and a correction. We found zero AI scholarships with a dated, open deadline that verifiably accept associate-degree students. Worse, last month's single pass is in doubt: for the DoD SMART Scholarship (still open, December 4, 5 p.m. ET), no source we could reach affirmatively includes associate-degree or community college students, and two restrict it to bachelor's and above or to a "university." The official site is JavaScript-gated and could not be read. Do not present SMART to students as confirmed-eligible until someone opens it in a browser. Read more →
AI Competitions
A $100,000 all-cash competition with a student-only prize lane — and a Monday deadline
All figures below are read off each competition's live main-page sidebar. Daylight saving ended Sunday, November 1, so every deadline from today is PST, and some sidebars still mislabel it.
Kiroween 2026 (Kiro, by AWS) is the best new item in weeks. $100,000, 100% cash, and the components sum exactly: $30,000 / $20,000 / $10,000, plus $10,000 Best Startup Project, four $5,000 category awards, and — the reason it leads — four Best Student Project awards at $1,500 each, a lane where a community college student competes against students rather than professional developers. Deadline December 4 @ 2:00pm PST; age of majority, no student requirement for the main pool so faculty and staff may enter; US not on the excluded list; 5,789 participants. Kiro subscriptions and blog-prize credits are stacked on top of the cash, not inside it. Read more →
Codegeist 2026 (Atlassian) is the bigger pool: $120,000, 100% cash, summing exactly across 27 cash winners — three at $15,000, three at $8,000, three at $5,000 and 24 category awards at $1,500. Deadline December 22 @ 12:00pm PST, no student requirement, 1,239 participants. Conference tickets are on top, travel not included.
Best odds for a student: DATANOVA Global Data-thon — students only, teams required, 64 participants against a $1,750 pool, closing November 10 @ 2:00pm PST (its sidebar reads "Nov 11 @ 3:30am IST," and the date moves back a day in conversion). Caveat: each prize is "a sum of cash and subscriptions" with no subscription value stated, so actual cash is below $1,750. GenHacks is smaller still at 34 participants, CAD 1,050 all cash, closing December 25 @ 9:00pm PST.
Closing soon, re-verified today. Codex x TwitchCon closes at 2:00pm PST today ($25,000 all cash, no student requirement, now 2,040 participants). Then Next Byte Hacks V5 Nov 7, Learn AR with SPECS Nov 9 @ 2:00pm PST ($2,500 all cash), Herhack Nov 10 @ 8:45pm PST, PayPal Nov 12 @ 12:00pm PST ($67,500 all cash, 26,648 entrants), UGREEN AI NAS Nov 15 @ 11:45pm PST ($46,000 all cash), Meta VR Start Nov 18 @ 12:00pm PST ($1,000,000, 100% cash, 4,821 entrants), LaunchPad Hacks and AccessED Hacks Nov 20, STEM2Connect Nov 21 @ 8:59pm PST. Nebius, OpenCV and GitLab have closed. Read more →
"$100k+ in prizes" that pays nothing, "$21,394 in cash" with almost no cash, and a correction on NASA
This is the part a student most needs. DSH Hacks V2 advertises "$100k+ in prizes!" in its own event title. Its sidebar says "5 non-cash prizes," and there is no cash prize of any kind. Add up every number the page publishes and you reach roughly $40,800 — all vendor credits, licences and gift cards, the bulk of it 100 winners of a $299 credit plan the page labels in capitals "THIS IS NOT A CASH PRIZE." Closes November 14 @ 11:45pm PST, 926 participants.
LarpHacks is worse in kind: a headline of "$21,394 in cash" of which $19,400 — 91% — is a sign-up prize described in its own text as AI vendor credits, and a "$500 cash" award whose description says $500 in Render credits. Its page also says the event has ended. Then the pattern worth naming: AccessED Hacks lists six awards totaling $225, five of which read "No cash or merchandise," and SafeSpace Hacks lists six totaling $280 on the same structure. Both are honest in saying so — but "6 prizes" is one cheque.
The correction we owe readers concerns NASA. We published the ORBIT Challenge as a rare explicit community college invitation. The official rules do say NASA "welcomes applications from accredited U.S.-based community colleges" — but they add a gate we missed: the lead institution must appear as an HBCU, TCU or Eligible Institution of Higher Education on the Department of Education's FY 2026 Eligibility Matrix. The invitation is conditional on that listing, and Fullerton's status has to be checked before anyone enters. Dates are now organizer-sourced and internally contradictory: the Timeline page says registration closes November 16 while the Rules page runs Phase 1 registration to December 14 — register by November 16. Phase 1 submissions close December 14 @ 12:00pm PST, the $500,000 purse does sum, and none of it pays out for the December submission. Stop citing the nasa.gov page, which describes a closed February 2026 cycle at $380,000.
Three smaller corrections. STEM2Connect now lists four non-cash prizes, not three — a .xyz domain award was added. Every participant count we published last week is stale, several badly: we printed Codex x TwitchCon at 587 and it is now 2,040, so we will date-stamp counts from here. And on AI film and art contests, a clean negative worth stating plainly: we reviewed roughly 26 free-to-enter contests with November and December deadlines and not one publishes a deadline with both a time and a time zone; several never say whether any of the prize is cash. None is usable. The nearest in-person event is SanD Hacks, La Jolla, November 14–15, about 90 miles down the I-5. Read more →
Embodied AI & Robotics
The FCC opened 50 MHz of cellular spectrum to drones, ending a ban that dated to 1991
The FCC voted 3–0 on Thursday, October 29 to adopt Report and Order FCC 26-72, opening 50 MHz of licensed, nationwide, low-band 800 MHz cellular spectrum to drones and ending an airborne prohibition in place since 1991 — roughly 35 years. The effective date follows Federal Register publication by 30 days; the draft set it there and the adopted order's date was not public as of the vote. Amazon, AT&T, Verizon, AUVSI and the Commercial Drone Alliance backed it. The order does not authorize beyond-visual-line-of-sight flight: FAA Part 108 remains unsigned and has not appeared in the Federal Register, against a statutory deadline of December 7, 2026 — now 35 days out.
For Fullerton College this is the most directly relevant robotics item of the week. The BS in Drones and Autonomous Systems started this fall, and this order is the communications layer underneath the entire BVLOS business case that program trains people to operate in. It also sets up a genuinely teachable regulatory contrast: the same agency is simultaneously restricting ground robots — mobile robots over 4.4 lb cannot get new equipment authorization under July's Public Notice DA-26-786 — and expanding drone spectrum. Curriculum hooks: a UAS spectrum and communications module, and Part 108 regulatory-readiness content. Job titles a program can credential for: UAS communications technician, BVLOS operations specialist, UAS regulatory compliance analyst.
Two follow-ons with dates. The FCC votes November 19 on further drone rules — 450 MHz air-ground data links including eVTOL, counter-drone detection radar at 24.45–24.65 GHz (detection and tracking only, not authority to disable), and expanded satellite access for moving platforms. And FEMA's $250 million counter-drone grant program closes November 16; we have not verified whether public higher education may apply, and at two weeks out that is worth a phone call rather than an assumption. Read more →
Humanoids became ordinary warehouse infrastructure, and Waymo's map now covers your service area
GXO and Agility Robotics expanded their partnership on October 26, adding sites in Statesville, North Carolina and Lebanon, Tennessee to the original Flowery Branch, Georgia operation, where Digit has moved more than 100,000 totes. Deployments begin early 2027 on Digit 4, migrating to Digit 5. Digit now has over 65,000 hours of real-world operation; RoboFab in Salem, Oregon is built for up to 10,000 robots a year. The structure is the curriculum hook: this is a multi-year Robots-as-a-Service agreement running on Agility Arc, a cloud fleet-management platform. These robots are managed, not merely installed — which is a job. Said plainly, the release contains no mention of retraining, job-role change or workforce impact, and the omission is itself the story for a CTE audience. Job titles: robot fleet operations technician, RaaS deployment technician, warehouse automation maintenance technician.
Agility also reported more than $300 million in committed multi-year orders for Digit 5 on October 26 — and its own release says this is potential value subject to milestones and "does not represent current-period revenue," a qualifier several outlets dropped. New customers named: Toyota Motor Manufacturing Canada and Schaeffler in automotive manufacturing. On October 30 the company announced an expected $465 million gross raise alongside its Nasdaq listing.
Closer to home: Waymo added nearly 250 square miles across five cities on October 27, and three are in California. Los Angeles extends through Manhattan Beach and Redondo Beach to Torrance; San Diego extends north to La Jolla, University City and Clairemont; the Bay Area reaches San Jose and Campbell with a BART connection at Berryessa. Waymo now runs in 15 metros at over 500,000 paid rides a week. Then on October 30, Waymo named the 2027 Toyota bZ4X as its next robotaxi — with Toyota manufacturing the vehicle and integrating Waymo's system at Takaoka in Japan, and the two companies continuing to explore Waymo technology in personally owned Toyota vehicles. For an automotive technology program that last thread is the one that matters: servicing a corporate robotaxi fleet and servicing a customer's own Level 4 car are different curricula. Read more →
This roundup was assembled with AI-assisted research by the EDUAIATLAS team. Spot something we missed? Let us know.