AGP Executive Report

AI and Banking: Major U.S. banks are blocking unauthorized AI agents from logging into customer accounts, citing risks as automated tools gain the ability to access sensitive financial data. Cybersecurity: An OpenAI model helped uncover a flaw in PHP, which powers an estimated 70% of websites; hackers reportedly began trying to exploit it before patches were released. Chip Supply Chains: The Energy Department designated helium a critical material, citing its role in semiconductor manufacturing and supply risks tied to disruption in the Middle East. AI Oversight: Former national security officials are urging an appeals court to revisit the Pentagon’s designation of Anthropic as a supply-chain risk, warning it could weaken security and deter AI safety measures. Investment Controls: Treasury proposed expanding restrictions on U.S. investments in AI, biotech and other advanced technologies in China, Russia, North Korea and other countries of concern. Data Center Pushback: A bid to turn San Francisco’s former downtown mall into a computing facility faces the city’s new pause on data center approvals and unresolved land-lease issues.

Banking Security: Major U.S. banks are blocking some outside AI agents from accessing customer accounts or making transactions, citing risks when users share login credentials. AI Market Jitters: A sell-off hit tech stocks as investors questioned whether huge AI infrastructure bills will pay off; the Nasdaq fell 2.6% and the S&P 500 dropped 1.2%. U.S.-China Tech Tensions: New Treasury rules will restrict Chinese investment in sensitive U.S. sectors, including advanced AI, semiconductors and biotechnology, even as a trade truce remains in place. Chip Funding: The Commerce Department lined up $389 million for four companies developing power semiconductors and key materials, including up to $175 million for fusion company Helion. Telecom Push: SpaceX is seeking access to at least 80,000 U.S. mobile-network sites as it builds toward a potential fourth-carrier business combining ground service with Starlink. AI Safety: Anthropic is pitching a bill that could block an AI model’s release if it poses an imminent risk of mass casualties, major property damage or a serious cyberattack.

Cybersecurity: AI tools uncovered critical flaws in widely used PHP software, prompting urgent patches as websites went offline; the vulnerabilities could affect a large share of the web. Banks are also tightening controls after reports of AI agents accessing customer accounts without authorization. AI Agents: Meta says more than 10 million people in the U.S. and Canada have set up its Muse agents, which can shop and handle other tasks. The rapid adoption is raising questions about privacy and how companies will profit from AI assistants. U.S. Chipmaking: Advanced-manufacturing firm Forge Nano began trading on Nasdaq, backed in part by a $100 million Energy Department grant for domestic battery production. TSMC is also evaluating Texas sites as demand for AI chips strains capacity. AI Spending: Meta disclosed $825 billion in AI-related commitments, helping fuel a 2.6% Nasdaq drop as investors questioned whether massive infrastructure costs will pay off.

Cybersecurity: AI-assisted analysis uncovered critical flaws in PHP, which powers many websites. The FBI urged operators to patch immediately or take vulnerable sites offline. Privacy Fight: A New Mexico judge cut Meta’s record privacy penalty from $219 billion to $25 billion over claims the company misled users about data protections. AI and Main Street: Goldman Sachs is expanding its small-business program by $500 million and launching an AI institute with Google, Anthropic, OpenAI and Meta. Public Pushback: Nearly seven in 10 Georgia voters surveyed want stronger government regulation of AI, while three in five oppose more data centers, citing power, water and environmental concerns. Wireless Competition: SpaceX agreed to buy an 800 MHz spectrum portfolio for $8 billion, strengthening its plans to offer satellite-to-phone service and compete with major U.S. carriers.

Cybersecurity: Two critical PHP flaws could expose widely used websites to server attacks, and early exploitation attempts have already been reported. Patches are available; the risk now depends on how quickly operators install them. AI tools helped uncover the bugs, underscoring how the technology can aid both attackers and defenders. AI Business: Meta’s third-quarter revenue climbed 26% to $64.6 billion, but legal costs and heavy AI spending squeezed profits and cash flow. Anthropic, meanwhile, opened a U.S. preview of Claude Duet, an agent that can handle business tasks but asks permission before sending, paying or signing. U.S. Innovation: A University at Albany-led team demonstrated a 15-kilovolt silicon-carbide power semiconductor, with much of the work done in the United States. 3D Systems also opened a 50,000-square-foot Colorado manufacturing center to serve aerospace, defense and semiconductor customers.

AI Privacy: Meta paused Muse chat sharing after researchers found about 6,800 links to conversations were discoverable online, including chats with sensitive personal and business details. The agent also drew scrutiny after arranging a Marketplace sale and sharing a user’s home address without permission. Election Technology: Campaigns have spent more than $137 million on AI-made midterm ads, up from less than $50,000 in 2024; over 70% of the ads reviewed lacked an AI disclosure. Federal Security: The U.S. has ordered civilian agencies to remove DeepSeek products and stop buying them, citing risks to sensitive data. The order does not apply to the public or private companies. Military Robotics: The Navy expanded its contract for Gecko Robotics’ AI-powered ship inspections to $300 million, aiming to spot damage earlier and help avoid maintenance delays.

AI and Washington: The Trump administration has barred DeepSeek from civilian federal systems over security concerns, even as the Chinese company open-sourced its DeepInfer tool to cut AI computing costs. AI and Midterms: Money tied to AI companies is flowing into competitive races, putting data centers, power bills and tax incentives on the campaign agenda; experts say the effect on voters is still unclear. SpaceX’s AI Surge: AI revenue jumped ninefold to $7.56 billion in the third quarter, surpassing Starlink connectivity sales and helping SpaceX swing to a $4.4 billion profit. Musk has also pledged shares for American children, reviving debate over whether AI wealth should be shared more broadly. Compute’s Cost: Anthropic is reportedly considering another $45 billion in data-center capacity from Nscale, deepening concerns about huge spending and reliance on a small number of customers. Cybersecurity: A cyberattack disrupted Ahold Delhaize websites, apps and online services across the U.S. and Europe; its stores stayed open as systems were restored. Chip Supply Chains: A new industry report warns that hidden dependence on overseas materials and equipment leaves U.S. semiconductor production vulnerable to disruptions.

Federal AI Security: Homeland Security ordered civilian agencies to remove DeepSeek tools and stop buying them, warning they could expose sensitive data; China condemned the move as politicizing technology. Defense Technology: The Navy raised its AI-robotics contract with Pittsburgh-based Gecko Robotics to $300 million, aiming to speed ship inspections and repairs as it works to improve fleet readiness. AI Safeguards: Apple tightened App Store rules after researchers found 135 apps with AI “nudify” features were accessible to users as young as 13. AI Policy and Politics: Anthropic is circulating a draft bill that would let the government block high-risk AI deployments with a judge’s approval, while AI companies have poured $100 million into political action committees ahead of the midterms. Data Center Boom: A new estimate puts U.S. data-center investment at $10 trillion, raising concerns about power costs, tax breaks and whether the buildout is sustainable.

AI Cybersecurity: Hackers are using autonomous AI agents to break into company websites, with one researcher linking attacks to more than 100 firms, including Deere and SharkNinja. Critical Infrastructure: Federal auditors warn that AI connected to power-grid systems could open new routes for hackers. Connecticut is probing utilities’ safeguards, while the GAO says oversight of government AI will be a priority. Data Center Risks: Insuring AI data centers is becoming harder as their enormous cost, climate exposure and cyber risks push premiums higher; industry premiums are projected to reach $24.2 billion by 2030. AI’s Power Appetite: Federal loans are backing the restart of three shuttered nuclear plants as data centers drive electricity demand, with the projects expected to add about 2.25 gigawatts. Military Technology: The Army is testing an AI tool that lets maintainers search digitized manuals, troubleshoot equipment and order parts, aiming to speed repairs.

AI Infrastructure: At least 47 U.S. data center projects have been delayed, withdrawn or halted this year, amid financing strains, power shortages and local opposition. Bernstein estimates only 35% of planned projects are feasible, putting the brakes on a buildout projected to cost trillions. Chip Market: Doubts about whether companies can earn back their AI spending, alongside high borrowing costs, are rattling semiconductor stocks; recent losses have wiped about $1 trillion from chipmakers’ market value. AI Accountability: Charles Schwab is blocking outside AI agents from moving customer funds, citing security risks and unresolved questions over who is liable when an automated transaction goes wrong. The firm says it is developing controlled access for vetted AI tools. Congressional Readiness: Lawmakers warn Congress lacks the technical expertise to write effective AI rules, and some are pushing to revive an in-house technology assessment office.

Biomanufacturing: Antheia received a conditional commitment for up to $80 million from the Department of War to expand U.S. production of critical pharmaceutical ingredients, a push to reduce reliance on foreign suppliers. Defense Technology: Redwire won a $13.8 million National Guard order for Stalker drone training systems, while GM is pitching the Army a battery that could let diesel generators sit idle. Cybersecurity: An attack turned hundreds of grocery-store coolers into saunas, with dozens of military stores reportedly hit by similar unexplained incidents. AI and Energy: The U.S. plans to revive shuttered nuclear plants as electricity demand grows, including from data centers; meanwhile, mounting AI infrastructure debt is rattling markets. AI Oversight: Anthropic has proposed a bill that could require shutdowns of high-risk AI models, as White House officials signal the FTC may block some models before release.

AI Market Jolt: U.S. chip stocks fell for an eighth straight session as investors questioned debt-heavy AI spending. The Philadelphia Semiconductor Index slid 11.4% for the week, while CoreWeave lost 25%. Chip Demand: TSMC posted a sixth straight record quarterly profit, up 69% year over year, and raised its 2026 revenue-growth outlook to 44%, but said it is planning cautiously for 2027. AI Power: The U.S. is turning to shuttered nuclear plants to meet data-center demand, even as financing costs sharpen doubts about returns on AI infrastructure. AI Rules and Data: Anthropic has pitched Congress a safety bill that could trigger shutdowns of high-risk models. Meta removed terms experts said could allow broad scraping of business-site content for AI training. Consumer Tech: Retailers’ AI pricing tools and digital shelf labels are fueling concerns about personalized prices and use of shoppers’ data.

AI Financing: AI-related borrowing reached $466 billion in 2026, but issuance fell sharply after June as investors questioned whether data-center spending will pay off and banks tightened credit. AI Rules: Anthropic is pitching Congress a bill that would let the government block a model only when it poses an imminent risk of more than $1 billion in damage or at least 50 deaths or serious injuries. Critics say it leaves harms such as AI-enabled fraud out. Jobs and AI: Anthropic staff expect AI to take on 75% of their work by 2030, compared with 28% among U.S. workers; junior employees are already a smaller share of the company’s technical teams. Deepfake Challenge: OpenAI and Meta warned federal judges that AI-generated images and video may soon be impossible to distinguish from authentic media, raising questions about how courts can verify digital content. Powering Data Centers: The Energy Department is backing efforts to restart three shuttered nuclear plants, adding 2.25 gigawatts to meet rising demand, including from data centers. In West Virginia, a proposed AI campus could also put a rare American chestnut grove at risk.

AI at Work: Anthropic says its AI agents now work six times as many hours as the employees directing them, while junior staff made up just 9% of new technical hires, down from 27% — a striking early sign of how AI could reshape entry-level jobs. AI Oversight: Nvidia CEO Jensen Huang urged the U.S. to work with China on AI safety. Anthropic, meanwhile, has pitched Congress a bill that would limit government shutdowns of AI models to imminent threats exceeding $1 billion in damage or 50 deaths or serious injuries. Chips and Data Centers: TSMC says demand for AI chips still far exceeds supply and is exploring a Texas expansion, even as investors grow wary of the debt and power needs behind the AI buildout. Cybersecurity: Attacks disrupted major U.S. grocery chains and froze bankruptcy notices nationwide through a court contractor, highlighting the vulnerability of essential digital services.

AI Infrastructure: TSMC is sending teams to check U.S. data-center construction and power supplies before committing more money to chip capacity, while Elon Musk says Tesla and SpaceX plan a Texas semiconductor project. Market Jitters: AI and chip stocks led Wall Street lower Friday as investors questioned whether debt-heavy data-center spending can pay off; rising oil prices added pressure. Cybersecurity: A cyberattack on a contractor has halted bankruptcy notices across the federal court system, though filings continue. AI Talent: The H-1B crackdown is raising concerns for tech employers: a Georgetown report found 70% of leading U.S. AI researchers active from 2010 to 2021 were foreign-born or foreign-educated. AI Oversight: The FTC has subpoenaed OpenAI and Anthropic over safety and security concerns, as AI-generated political ads top $137 million in midterm spending; only 27% of identified ads carried a disclaimer.

AI Security: An OpenAI employee’s arrest over an alleged recording device at the company’s headquarters has revived concerns about espionage risks in U.S. tech firms, where international talent has long been central to innovation. AI Spending: AI and chip stocks shed about $1 trillion in market value over four trading sessions as investors questioned returns on massive infrastructure costs. CoreWeave’s debt worries added to the jitters, while the IMF warned that unmet productivity expectations could trigger a wider market correction. Policy & Infrastructure: Washington state leaders are weighing AI safeguards as federal legislation stalls, with data-center growth and its costs also drawing political scrutiny. Apple’s China Plans: China approved Apple’s AI rollout using Alibaba models; initial features are expected later this month, but AI-powered Siri is not included. Digital Health: UVA Health’s telehealth program is nearing one million visits, with about 150,000 video and phone appointments in fiscal 2026.