
Your SOC 2, handled end to end with AI agents.
Enterprise buyers won’t put your product near their customer data without a SOC 2 report. Twenty-one state privacy laws now hold them accountable for the vendors they share data with, so their diligence lands on you.
Sprinto gets you audit ready in 14 days, across three working sessions. AI agents connect your stack, collect the evidence auditors ask for, and close the gaps as they appear. You approve, they execute.
Sprinto also answers the security questionnaires your buyers send, using the same evidence base, so security reviews stop holding up deals.
No compliance hire. No consultant. Your auditor signs off.
⚖️ POLICY & REGULATION
A Conservative AI Safety Group Puts Out An "America First" Platform: Mandatory Testing, Incident Reporting, Developer Liability, A Ban On Recursive Self-Improvement And An H-1B Pause: Humans First, chaired by Tea Party veteran Amy Kremer and run by 2020 Trump campaign adviser Russ Walker, shared its "America First AI Policy" memo with Semafor, which reported it Thursday, Oct. 15. The eight planks: tighter export controls against China and other adversaries, mandatory independent testing for large AI companies, a prohibition on developing practices such as recursive self-improvement, a ban on federal use of Flock-style cameras, public reporting of AI incidents, liability for frontier developers whose conduct is reckless, strict rules for AI used by children, and a pause on H-1B visas. Walker said conservatives should lead on an AI policy centered on the American people; Kremer is running a bus tour billed as the "AI Data Center Revolt." Testing, incident reporting and liability overlap with Cantwell's framework and the Hawley-Murphy liability bill, which puts a Republican-branded group to the left of the administration's light-touch line and of most Hill Republicans on the same three questions. Link
GAO Finds The Pentagon's Responsible AI Council Did Not Meet For Nine Months, Has No Charter And A Division Staffed Three Of Ten; DOD Agrees To Fix Both: In a report published Wednesday, Oct. 14 (GAO-27-107837), the Government Accountability Office says the Chief Digital and AI Office's Responsible AI Working Council did not meet from late 2024 until August 2025, that officials from four of the ten components surveyed reported implementation problems during the gap, and that the council still has no charter defining roles or meeting frequency, so it is at risk of lapsing again; as of February 2026 only three of the ten authorized positions in the Responsible AI Division were filled. Six of ten components said shared resources and best practices would help. GAO recommends the Secretary of Defense have the CDAO issue a charter and address the division's workforce, and DOD concurred with both; it also asks Congress to consider directing DOD to act on its older open recommendations on AI collaboration and workforce. The report notes the January 2026 Secretary of Defense AI strategy memo redefined Responsible AI as using models free of usage-policy constraints that may limit lawful military applications, which is the policy the under-staffed division is now meant to implement. Link
Five Fed Districts Report Firms Cutting Staff By Attrition Or Delaying Hiring After AI Investments, In A Beige Book That Otherwise Finds Flat Employment: The Federal Reserve's October Beige Book, released Wednesday, Oct. 14, says five of the twelve districts reported that some firms cut staffing through attrition, delayed hiring or expected less job creation "after increasing productivity by investing in AI," while most districts described little or no net change in overall employment. The report is anecdotal by design: it gives no count of jobs affected and no estimate of AI's effect on national employment, and the national summary does not name the five districts. In July's edition contacts described AI mainly as a capital investment with headcount held steady, so this is the first time the national summary ties AI spending to fewer jobs, and it lands a day before New Jersey's A 5676 was introduced to tax AI compute at 10 percent for a workforce transition fund (The Docket, below). Link
AI Safety Donors Have Put About $48 Million Into The Midterms, Axios Finds, Against $44.8 Million Paid Out By The Industry's Leading The Future PAC Through June: An analysis shared first with Axios and reported Wednesday, Oct. 14, counts roughly $48 million from AI safety groups and individual donors across Democratic and Republican races this cycle, including contributions from individual researchers at OpenAI and Redwood Research. Guardrails Alliance, the tech-worker-backed group, has raised $11.7 million across a super PAC and a 501(c)(4), though its super PAC reported only $741,778 in receipts through June 30 in FEC filings; it launched in June with about $5 million and a $15 million goal. On the other side, Leading the Future, the industry-backed super PAC, had paid out $44.8 million and held $31 million in cash through June 30. Guardrails executive director Shaunna Thomas has framed the fight in partisan terms, and some Democrats argue that opposing their own candidates could elect Republicans less inclined to regulate AI. The analysis is Axios's own; the totals have not been reconciled against OpenSecrets or FEC filings. Link
STATES
Virginia Heads Into January With Four Data Center Moratorium Vehicles: Perry's Three-Year Pause, Roem's Queue-Linked Bill, Shin's Carried-Over House Bill And Sturtevant's Indefinite Version: Virginia Mercury reported Friday, Oct. 16, that Sen. Danica Roem (D-Prince William), a co-patron of Sen. Russet Perry's SB 843, is also filing her own moratorium bill that would block new statewide data center approvals until the large-load customers in Dominion Energy's interconnection queue are energized and the infrastructure to serve them is built, and Sen. Glen Sturtevant (R-Chesterfield) plans to file a version he wants to be indefinite, alongside a push to reopen the data center sales and use tax exemption and a call for a special session. Del. Irene Shin's House bill, carried over from 2026, would bar localities from approving zoning and permit requests until pending interconnection requests for data center customers are fulfilled or July 2028, whichever comes first. SB 843 itself, released Oct. 13, would bar localities, DEQ and the SCC from approving data center permits, distribution lines and on-site power from July 2027 through July 2030, ban NDAs for new projects and order noise regulations. "Local or state moratoriums could deprive communities of the opportunity to compete for investment and jobs," said the Data Center Coalition's Nicole Riley; IBEW Local 26's Don Slaiman said "I don't think a moratorium is gonna pass because you're taking out your tax base." Gov. Spanberger's September framework includes no moratorium. The session convenes Jan. 13. Link
Paxton Sends Water-Use Demand Letters To 156 Data Center Developments In 32 Texas Counties: The Texas Attorney General's office said Thursday, Oct. 15, that it sent letters to 156 data center developments in 32 counties, Bexar, Dallas, Harris, Travis, Tarrant, Williamson, Collin, Denton, Tom Green, Midland, Lubbock, El Paso and Webb among them, notifying them that they must provide "current and accurate information about their water usage." The letters follow the investigation Paxton opened Sept. 24 into developments that have not answered the Texas Water Development Board's water-use survey, which Gov. Abbott ordered the board to enforce after reporting that fewer than a third of the companies required to respond had done so; non-respondents can lose eligibility for board financial assistance and water rights permits. The office released no copies of the letters and set no public deadline; KBTX's request for them went unanswered. It is the water-side companion to the TCEQ permit halt whose compliance update is due to Abbott on Monday, Oct. 19. Link
Johnson County, Iowa, Extends Its Data Center Moratorium Nine Months To August 2027 As Nearly Half Of Iowa's Counties Now Have A Pause: The Johnson County Board of Supervisors voted unanimously Thursday, Oct. 15, to extend its moratorium on data center development through Aug. 8, 2027, with the option to end it sooner if staff research and the public hearing process finish first; planning staff will draft ordinance amendments covering water impacts, parking and road wear during construction and operation, review other states' rules and work with the citizen advisory commission and the Planning and Zoning Commission. Statewide, nearly half of Iowa's 99 counties have passed some form of data center moratorium, most running six months to a year, according to the Iowa AI Coalition's county tally reported by WHO 13 the same day; the same report notes Microsoft's purchase of 600 acres near Woodward for a potential project. Link
Aurora's Mayor Calls A Monday Special Meeting On A Data Center Application Moratorium He Blocked In The Spring, Guilford County Punts To Nov. 5, Hawaii Island's Second Planning Commission Backs The Ban, New Castle County Adds An Infrastructure Test: Aurora, Colo., Mayor Mike Coffman has called a special council meeting for 6 p.m. Monday, Oct. 19, on Councilmember Amy Wiles's proposal to stop accepting data center applications until the city adopts regulations, writing that "we cannot be blindsided by another Trident Data Center disaster"; Coffman cast the tie-breaking vote against a six-month moratorium earlier this year, the term of the new pause is undecided, and the already-accepted 334-MW Trident application would likely be unaffected. Guilford County, N.C., commissioners held their public hearing Thursday, Oct. 15, on a 180-day moratorium on new or expanded data centers and crypto mining in the unincorporated county, to April 12, 2027, then postponed the vote to Nov. 5 on Commissioner Alan Perdue's motion; residents asked for 32 months, and Chairman Skip Alston said "we're gonna get it right the first time, and not in a rush." Hawaii County's Leeward Planning Commission voted unanimously Oct. 15 to send Councilwoman Heather Kimball's Bill 170, which removes data centers as a permitted use countywide, to the council with a favorable recommendation, two weeks after the Windward commission voted it unfavorable; the Planning Department warned a ban could push developers onto state land the county cannot control. New Castle County, Del., council passed its second data center ordinance Tuesday, Oct. 13, requiring any large energy-use facility to show adequate electricity and transmission capacity before approval, not retroactive to projects in the pipeline, parallel to the state's June energy-service-agreement law; it awaits County Executive Marcus Henry's signature. Link
🛡️ SAFETY & SECURITY
Anthropic Gives Six METR Staff Employee-Level Access, With Publication Rights It Says It Cannot Veto, And Sets Out A Three-Tier System For Outside Evaluators: In "An update on independent evaluators," posted Thursday, Oct. 15, Anthropic said six employees of the nonprofit METR now have company laptops, office space and internal tools roughly comparable to its own risk assessment teams, with customer data, legally privileged information and model architecture off limits; Anthropic says it cannot veto METR's findings on risks, incidents and practices, can require only narrowly defined redactions, and METR may publicly challenge a redaction without revealing what was withheld. The post describes three tiers: narrowly scoped investigations where publication needs approval (SecureBio, reviewing the chemical and biological chapter of the August risk report), broader access with publication subject to challengeable redactions, and METR's employee-like tier; Faculty, Accenture's AI unit, is a partner and talks with Hugging Face are under way. METR said it "will openly describe the scope, access, and other limitations of our work," and a person familiar said Anthropic had become "more open in negotiations." It answers the Reuters report two days earlier that both labs had sought control over what evaluators publish, and it arrives the week the FTC said it would compel METR executives to testify. Link
Gates Is Preparing A Larger Personal AI Safety Campaign Aimed At CEOs And Possibly Funded With His Own Money, Axios Reports: Axios reported Thursday, Oct. 15, that Bill Gates is preparing a much larger AI safety effort that would press business leaders directly and could commit a significant amount of his own money, after consulting AI executives and safety advocates including Anthropic's Dario Amodei; people close to him say he has grown dismayed by the escalation of model capabilities, the summer's security incidents and what he sees as insufficient regulation. "I can reach CEOs on this better than you can," he reportedly told one adviser. The report gives no funding figure and no timetable. Gates called for congressional legislation and said self-regulation is not enough on Meet the Press on Sept. 27, a position President Trump has dismissed as a "hoax." For the labs, it adds a second outside voice pressing them on the voluntary accord's four commitments while the Super Intelligence Force audits them. Link
💰 POWER & CAPITAL
NERC's Data Center Reliability Standards Clear Their Final Ballot: 50 MW And Up At 100 kV, Board Vote This Quarter, FERC Filing Due Dec. 31: NERC said Thursday, Oct. 15, that stakeholders approved the computational-load reliability standards on a final ballot. They apply to AI data centers, crypto mines and other computational loads of 50 MW or more supplied through equipment connected at 100 kV or higher, and cover modeling data, verification and reporting for planning cases, the site data transmission operators, balancing authorities and reliability coordinators need to plan and monitor operation, predefined communication protocols for prompt action during events, interconnection studies, load protection coordination and disturbance monitoring. The package goes to the Board of Trustees, which NERC expects to take it up in the fourth quarter, and then to FERC, whose July 16 order requires the first computational-load standards to be filed by Dec. 31; on Oct. 8 the board approved Rules of Procedure changes creating Computational Load Owner and Operator as registered entity types, which is how a customer becomes a NERC-regulated party for the first time. The standards grew out of the May Level 3 alert after sites dropped or oscillated load, and NERC projects summer peak demand to grow 224 GW over ten years. Link
Twelve Republican Attorneys General Ask FERC To Approve The Green Chile Pipeline For Oracle's Project Jupiter And Reject New Mexico's Bid To Review The Data Center With It: In a letter to FERC on Wednesday, Oct. 14, reported by Bloomberg, the attorneys general of West Virginia, Alaska, Arkansas, Florida, Georgia, Iowa, Louisiana, Mississippi, Montana, Nebraska, Ohio and South Carolina asked the commission to authorize Energy Transfer's 17.8-mile Green Chile pipeline in Doña Ana County, which would carry up to 400 million cubic feet of gas a day off the Transwestern system to the generation serving Oracle's campus, and argued that FERC has no authority over data centers and so cannot expand its environmental review to include the campus, as New Mexico Attorney General Raúl Torrez asked in his Oct. 5 filing calling the review inadequate. The states warn that FERC's approach could be applied to gas-fed data center projects everywhere, which is the point: the question of whether the load behind a pipeline is part of the project is the one the Sierra Club and the Center for Biological Diversity have raised too. The state air permit for the campus's fuel cells is on a separate clock, with the Environment Department's decision due by Nov. 23. Link
Wisconsin's PSC Approves Two Invenergy Gas Peakers Totaling 1.5 GW For Data Centers 3-0, With Commissioners Saying State Law Left Them No Way To Weigh Need And Calling It "Ripe For Legislative Action": The three-member Public Service Commission unanimously approved Invenergy's Foundry Ridge Energy Center in Walworth County and Red Oak Ridge Energy Center in the Town of Paris, Kenosha County, on Thursday, Oct. 15; together they exceed 1.5 gigawatts, would run on peak-demand days, and would serve the data center campuses in Mount Pleasant and Port Washington, with Foundry Ridge in service in late 2028 and Red Oak Ridge between late 2029 and early 2030. We Energies has a pending filing to buy power from both for a combined $2.3 billion, which is where need is weighed: under Wisconsin law a merchant developer's plant is approved without the public-need test a utility plant faces, and air permits belong to the DNR. Commissioner Kristy Nieto called it "gas for AI" and questioned whether existing law gives regulators the tools to evaluate it; Chair Summer Strand called the merchant route "a two-step, piecemeal, often whack-a-mole approach" and said "these deficiencies in the current framework are ripe for legislative action." Healthy Climate Wisconsin and residents opposed the plants. Link
New York City's Comptroller Asks Anthropic To Delay Its IPO Until It Sets Out Audits, Controls And A Timeline On Safety And Governance: In a letter to Dario Amodei that CNBC reported Thursday, Oct. 15, and Reuters reviewed the same day, Comptroller Mark Levine, whose office oversees city pension funds with more than $325 billion in assets and about $7 million of exposure to Anthropic through private equity, urged the company to delay its U.S. IPO until it addresses concerns about safety, governance and the pace of development, and asked how it will implement audits and controls, what they cover, on what timeline, and whether they answer Amodei's own call to slow model development; the company should "focus its energies on addressing these unprecedented risks," he wrote. Reuters' sources say Anthropic told prospective investors that third-quarter revenue rose 45 percent from the prior quarter and that it is targeting a $2 trillion valuation; the company did not respond. A pension fiduciary asking an issuer to hold its offering is a new pressure point, and it comes the same week the FTC's demands and the state suits were filed and the day Anthropic published its evaluator-access terms (Safety, above). Link
📡 PRODUCTS & SIGNALS
Meta Releases Muse Spark Open, An 816-Billion-Parameter Open-Weight Model It Says Cleared Its Scaling Framework, Under A License With Limited Commercial Terms: Announced Thursday, Oct. 15, Muse Spark Open is a natively multimodal model with 816 billion total and 42 billion active parameters and a one-million-token context window, built for coding and agentic work; Meta says it runs on two Apple M5 Ultra Mac Studios through the MLX framework, each with up to 512 GB of unified memory, and on AMD, Intel and Nvidia hardware through SGLang, vLLM and llama.cpp. Weights are on Hugging Face, Amazon SageMaker, Google Cloud Model Garden and Microsoft Foundry under the Muse Model License, which Meta describes as having limited commercial terms that still let many developers use it free; APIs run through Baseten, Fireworks, Groq, Modal, Nebius, OpenRouter and Together. Meta says the model received additional safety post-training because of open-weight considerations and "met the thresholds our framework requires for release" under its Superintelligence Scaling Framework, with a full evaluation report promised "soon." Meta calls it the most capable U.S. open-weight model to date, and it is the model class whose self-replication risk Anthropic's red team wrote up Oct. 12; Pennsylvania's HB 2800, heard Monday, covers any developer training above 10^26 operations whether the weights are open or not. Link
Amazon Launches Two Always-Listening Bee Wearables At $74.99 And $99.99 With A $11.99 Monthly Tier, Keeping Transcripts But Not Audio: On Thursday, Oct. 15, Amazon introduced the Bee Frontier, a wrist or clip-on recorder available now at $74.99, and the Bee Eclipse, a magnetic phone attachment at $99.99 listed as coming soon, the first products since it bought Bee in 2025. Free users get conversation summaries and recaps retained for 14 days; Bee Pro at $11.99 a month adds further summaries, suggested tasks and integrations with Alexa+ and Amazon shopping. Amazon says users control when the devices listen, can delete conversations, and that it does not store audio recordings. The devices record other people's speech by default, which is the fact pattern behind the all-party-consent states' wiretap laws and the ambient-recording disclosure laws now being written for clinical settings. Link
📜 DISPATCH PRO
The working layer under this briefing is The Docket, in Dispatch Pro
What moved in all 50 states since the last issue, which AI bills actually matter this week, and exactly what they would make you do, by when: obligations, deadlines and penalties pulled from the bill text and verified, similar bills in other states, every hearing and effective date we have reported, plus a Docket Room Pro seat ($49/mo on its own).
Today for Pro readers: 9 AI bills moved since the last issue, 1 rated Act; this week's three; The Bill on NH HB 1506 with 3 obligations, deadlines and penalties pulled from the text; 1 more AI hearing this week and 45 dated actions in the next 30 days, plus the full Deadline Board and the full Moratorium Tracker, Friday only.
$29/mo or $290/yr after the trial, less than the Docket Room seat alone. Cancel anytime. Team licenses on request.
Digitize Dispatch is read by lobbyists, government affairs, policy, and regulatory professionals tracking AI legislation, regulation, and policy developments. Put your brand in front of the people shaping how organizations respond to AI policy.
Interested in advertising with us or have any feedback? Send us an email.



