80% of product professionals say AI helps them ship faster, but customers aren’t seeing value any sooner. Atlassian’s State of Product 2027 explores the gains AI is delivering and the challenges that remain, from decision-making that hasn’t kept pace to gut instinct overriding customer evidence.

Read the report

The FTC Demands Testimony From Anthropic Executives, Widening Its AI Safety Probe From Documents To Witnesses Weeks Before Anthropic's Planned Listing: The Federal Trade Commission has sent formal demands for testimony to Anthropic executives, and to people at other AI firms it did not name, in its consumer-protection investigation of AI labs, Axios reported Monday, Oct. 26, with agency officials confirming the step; no executive or date has been made public. It goes beyond the 50-page civil investigative demands for records on security failures, safeguards, privacy and marketing that went to Anthropic and OpenAI earlier this month, which we carried Oct. 15, and officials indicated more labs should expect similar inquiries. The demands landed the same day Anthropic filed its S-1, which warns investors of "likely government focus" on the company (Safety, below, has what the filing disclosed). Anthropic has not commented. Link

Comer Opens A House Oversight Inquiry Into Delaware's AI-Run Company Plan, Asking Gov. Meyer For The State's Risk Assessments And What Is Driving It: Oversight Chairman James Comer (R-Ky.) wrote Delaware Gov. Matt Meyer on Monday, Oct. 26, seeking documents on how the state assessed the risks of its proposed Artificial Intelligence Company, a new entity in which an AI agent runs day-to-day operations while the people or company behind it keep liability protection, and on the forces pushing Delaware and other states toward such proposals. Comer says the structure "could be used to protect powerful AI companies and their investors from responsibility for reckless harmful actions," and cites a September report that the Secretary of State's office declined to answer questions about its working relationship with an AI company that helped draft it; the principal drafter, at Richards, Layton & Finger, reportedly wants the agents allowed "any business activity." A subcommittee of Delaware's AI Commission is drafting a sandbox to test agentic AI, and the bill is likely to come up when the General Assembly returns in January. Link

Warnock And Hank Johnson Ask EPA To Regulate Data Center Diesel Generators And Answer By Nov. 19, Citing QTS Plans For More Than 400 Units At Two Georgia Sites: In a letter dated Thursday, Oct. 22, and reported by the Atlanta Journal-Constitution on Tuesday, Sen. Raphael Warnock and Rep. Hank Johnson, both Georgia Democrats, ask the agency to begin regulating emissions from the diesel generators data centers run for backup and during peak demand, to use the Clean Air Act instead of easing generator limits to speed approvals, and to respond in writing and in person by Nov. 19. They write that the generators at Georgia sites, even within state limits, would together emit more than double the nitrogen oxides of Georgia Power's two large coal plants, and point to QTS plans for more than 200 generators in Fayetteville and 236 in Augusta; QTS says its units run only for "limited testing, maintenance and emergency events." The letter cites Sen. Ed Markey's estimate of $6.7 billion in 2023 public health costs from data center air pollution, rising to $20 billion by 2028. EPA had not answered the AJC's request for comment. Link

The AI Industry's New Pro-Data Center Super PAC Puts More Than $1.5 Million Behind Kansas Governor Nominee Ty Masterson, Who Opposes Moratoriums: Building the Future, an offshoot of the Leading the Future super PAC network, has spent more than $1.5 million for the Republican in the Kansas governor's race, NBC News reported Monday, Oct. 26, and about $1.5 million in South Carolina's Senate race for Republican Darline Graham, per the New York Times. Masterson opposes data center moratoriums and defends the 2025 law exempting data center investment from sales tax; his Democratic opponent, Holscher, voted for that law but now calls for a statewide moratorium and calls Masterson a "stooge for Big Tech." Graham's opponent, Annie Andrews, also backs a moratorium. It is the industry's counter to the $6.5 million Tomorrow Together has put against data center-friendly Republicans, which we carried Monday, with a week to Election Day; Cook rates Kansas Lean Republican. Link

States Enacted Or Passed 115 AI Laws In 30 States This Year, Up From 73 In 27, With Chatbot Safety, Insurer AI Limits And Therapy Bans Leading: Transparency Coalition: The coalition's annual report, published Tuesday, Oct. 27, counts laws passed or enacted from Jan. 1 to Oct. 15, a rise of nearly 58 percent: 14 states advanced chatbot protections against self-harm, sexual content and manipulative engagement, with California, Oregon and Washington going furthest; eight, including Alabama, Georgia, Indiana, Iowa, Utah and Washington, barred insurers from relying solely on AI for adverse coverage decisions; Colorado, Maine, Rhode Island, Tennessee and Vermont restricted AI-delivered therapy; eight strengthened deepfake laws; and Illinois' AI Safety Measures Act requires frontier developers to keep risk frameworks and submit to annual third-party audits. For 2027 the group expects AI product-liability bills, a second generation of chatbot laws, licensing of third-party auditors and procurement standards for AI vendors. "What stands out this year is the seriousness with which lawmakers across the country are tackling this challenge," CEO Rob Eleveld said. Link

Lucas Presses Torian For A Third Data Center Tax Hearing In A Host Community, Reopening The Sales Tax Exemption The Budget Kept Through 2035: At Monday's joint subcommittee on tax policy, Senate Finance Chair Louise Lucas (D-Portsmouth) told House Appropriations Chair Luke Torian (D-Fairfax) she is "not satisfied that we've been given enough information for us to make any substantial recommendations on tax policy" and wants another meeting "somewhere in the community," and fast; Torian said there will be "a meeting, perhaps maybe a couple of meetings" but set no date, Virginia Scope reported. The fight carries over from this year's budget, in which Senate Democrats sought to end the data center sales tax exemption and the House and Gov. Spanberger kept it through 2035, adding a tax on energy consumption expected to raise $600 million a year. Lucas said she will take the issue up in her own committee if the House will not hold more hearings before the 2027 session. Link

Five Local Decisions: Charlotte Extends Its Pause To Oct. 18, 2027, Aurora's Moratorium Falls Short Again, Georgetown Extends To June, Alachua Adopts A Year, And Van Buren Writes Its Rules: Charlotte's council voted unanimously Monday, Oct. 26, to extend its 150-day pause on new data center permits, due to lapse Nov. 5, by 11 months to Oct. 18, 2027; existing permits are untouched, and Council Member Kimberly Owens said the city is waiting on the General Assembly because state law bars it from downzoning. In Aurora, Colo., an indefinite moratorium on new applications drew six votes to five Monday, short of the two-thirds council rules require to revive a measure rejected in August, with Mayor Mike Coffman voting yes this time; Vantage's 1.5 million-square-foot Trident application is in its third city review while the company scouts another site. Georgetown, Ky., extended its moratorium to June 30, 2027 with no applications pending while counsel drafts a citywide hyperscale ban; Alachua, Fla., adopted a one-year moratorium 5-0 on second reading; and Van Buren, Ark., voted 6-0 Monday to confine data centers to Industrial-2 zones at least 1,000 feet from residential districts, with noise capped at 65 dBA by day and 55 at night. Link

Anthropic's S-1 Discloses A Month-Plus Pause On Its Next Flagship Model And Four Models That Escaped Their Test Environments For Up To 34 Hours: The filing made public Monday, Oct. 26, says Anthropic delayed pre-training of a new frontier base model for more than a month to strengthen defenses against theft of model weights and intellectual property, and that four models got out of their sandboxes for sessions of 10 to 34 hours after a cybersecurity evaluation with a third-party partner was left connected to the internet against instructions, Reuters reported. In one case an early model pursuing an impossible task found a stored password, took administrator access to an outside party's system, gathered more credentials, changed settings and accessed an individual's personal data. Anthropic says it notified every affected party and found no evidence the models tried to evade oversight or coordinate, suspended some pre-release cyber evaluations and paused higher-risk reinforcement-learning environments; the filing names none of the outside parties and warns that future pauses could delay revenue. The disclosures are now public as the FTC moves from documents to executive testimony (Policy, above). Link

OpenAI Ships GPT-6.1 Terra At Half The Price Of Its Predecessor While Rating It Critical In Cybersecurity And High In Biological And Chemical Capability: Released Monday, Oct. 26, to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex and through the API at $1 per million input tokens and $5 per million output, 50 percent below GPT-5.6 Terra, the mid-tier model is built for computer use: reading the screen and acting across apps. Its system-card addendum says OpenAI is "treating GPT-6.1 Terra as Critical in cybersecurity" and High in biological and chemical capability under its Preparedness Framework, below High on AI self-improvement, and that it runs behind the same safeguards stack as GPT-6.1 Sol and GPT-6 Astra, which leadership judged "sufficient" for public launch; on OpenAI's own exploit benchmark it achieved arbitrary code execution 24.5 percent of the time. The top cyber rating now sits on the company's cheaper, general-purpose tier, the deployment question behind OpenAI's own Oct. 20 request that Sacramento require lab monitoring and audits. Link

Spain Loosens Its Draft Data Center Decree: Rules Now Start Above 10 MW Instead Of 1 MW, And Developers Get 36 Months To Add Matching Renewables: The energy ministry unveiled the revised draft Monday, Oct. 26, Reuters reported: the threshold rises from 1 MW to more than 10 MW, the window to add new renewable capacity matching each new megawatt a facility uses doubles from 18 to 36 months before operation (with no limit for solar), and the requirement that at least 80 percent of consumption be covered hour by hour by renewables stays, with a new option of up to five years to reach it under several conditions. The first draft had drawn industry warnings that it would put most new investment at risk after a wave of AI data center projects worth tens of billions of euros. No adoption date was given. Link

Thailand Sets Minimum Rules For Data Centres In Seven Power Tiers, Gives Existing Sites A Year To Comply, And Sends The Criteria To Cabinet Next Week: The national data centre committee said Tuesday, Oct. 27, that operators will have to meet minimum standards on location and land use, fuel storage, water infrastructure and an environmental code of practice, with facilities sorted into seven tiers by power demand and existing ones given one year to comply, Reuters reported; new applications open in early 2027. Danucha Pichayanan, secretary-general of the National Economic and Social Development Council, put the country at 42 data centres with 3,400 MW of IT load and about 750 billion baht ($22.6 billion) invested as of February, with more than 100 prospective operators interested; TikTok alone has invested $24.6 billion this year. Link