AGP Executive Report

AI Cybersecurity Race: Chinese startup Z.ai says its GLM-5.2 can match Anthropic’s Mythos for finding software security bugs, narrowing the gap as open models raise both defense and misuse concerns. US AI Access Controls: Anthropic received US authorization for limited access to Mythos 5 for select American cybersecurity firms, after Commerce ordered a cutoff over safeguard vulnerabilities. Digital Lending Crackdown: Nigeria’s FCCPC denied claims it approved 48 new digital loan apps, citing a Federal High Court order that suspends the 2025 lending regulations. Digital Identity Overhaul: Nigeria’s Tinubu signed the NIMC Act 2026, expanding the commission’s role as a root certification authority for secure digital identity and trusted transactions. Software Reliability in the Real World: Hyundai recalled 96,000+ Tucson vehicles over a software issue that can blank the instrument panel display, while EPFO’s portal upgrade temporarily disrupted claims and e-passbook services. Tech Policy for Growth: Philippines PEZA plans to ask Marcos to lift the Metro Manila ecozone moratorium, aiming to unlock more IT-BPM investment. Hardware & Computing Culture: A homebrew Apple II Plus clone rebuilds the classic computer in hardware (no emulation), and a new hyperscale data center debate highlights long-term land and power impacts.

Semiconductor & AI Expansion: Micron will open an Atlanta memory-chip design center, aiming to tap Georgia Tech talent and add R&D capacity. Regional Tech Growth: India and the US are moving from talks to projects in semiconductors, AI, quantum tech, and critical minerals, with private firms expected to drive outcomes. Cybersecurity Model Access: Anthropic is partially restoring access to Claude Mythos 5 for selected US critical-infrastructure organizations after earlier government limits. AI “Bug-Finding” Race: Chinese firms 360 Security and Zhipu AI claim their AI vulnerability tools are now comparable to Anthropic’s Mythos, intensifying AI-for-cyber competition. Enterprise Software Costs: Microsoft’s July price increases are being framed as an “AI tax,” with some packages rising up to 33% and hitting frontline-heavy businesses hardest. Payments & Platform Rules: NPCI says it plans a 30% cap on any single UPI app’s market share by end-2026, plus a push toward a 1B daily UPI target. Fraud Watch: Researchers warn of fake GTA 6 “early access” sites taking crypto payments with no game delivered. Public Sector IT: Taiwan bus drivers in some cases report digital punch clocks malfunctioning with dashboard software, triggering operational complaints. Education & Training: Rogers State University plans a $35M Robson Center for Science and Technology, including cybersecurity and robotics programs.

AI Governance Shake-Up: OpenAI says it’s restricting release of GPT-5.6 Sol to Trump-approved customers, while Anthropic’s Mythos 5 gets a limited green light for trusted cyber defenders after earlier blocks. Cybersecurity Policy Push: Bangladesh’s Cyber Security (Amendment) Bill would force platforms like Meta to remove harmful AI-generated content within set timelines, with added penalties and powers for regulators. Software Reliability in Public Transit: OC Transpo reports an overnight software issue slowed charging for electric buses, delaying Saturday service and triggering garage shuffles. Anti-Cheat Change in Gaming: Riot updates Vanguard so it won’t start on PC boot; it runs only when launching supported Riot games. Hardware & Costs: Apple raises MacBook, iPad, and Apple TV prices due to surging memory/storage costs tied to AI data centers. India Digital Shift: India marks Digital India’s 11 years with a new push toward AI and semiconductor manufacturing, including a shared compute push backed by 45,000 GPUs. Open-Source Picks: Android users get a roundup of open-source utilities and trending free apps to replace defaults.

AI Model Access Controls: The US partially restores Anthropic’s Mythos 5 for about 100 vetted cybersecurity and critical-infrastructure organizations, while keeping Fable 5 largely blocked—another sign that frontier AI releases are being gated by national security reviews. OpenAI Model Rollout: OpenAI launches the GPT-5.6 family (Sol, Terra, Luna) in a limited preview for trusted partners, with broader availability planned as regulators scrutinize new capabilities. Vehicle Cybersecurity Rules: India’s MoRTH draft would make cybersecurity (AIS-189) and software-update management (AIS-190) mandatory for multiple vehicle categories, starting a phased rollout from October 2026. Cyber Fraud in Banking: Kerala police investigate a scam where forged court orders and fake police documents were used to defreeze six accounts and move funds. Public Sector AI Governance: UN officials and civil servants discuss balancing AI efficiency with transparency, accountability, and meaningful human involvement. Cyber Skills & Training: CyberWyoming rebrands its small-business cybersecurity program as CyberUSA to scale practical training nationally. Digital Governance Push (Pakistan): Pakistan’s proposed National Data Governance Policy would empower the Pakistan Digital Authority with audit and enforcement powers across federal bodies.

AI Governance & Cybersecurity: OpenAI says it’s limiting GPT-5.6 Sol to “trusted partners” approved by the Trump administration after a cybersecurity review, echoing broader government pressure on frontier AI labs. Defense Software: DISA is lining up more Joint Enterprise License Agreements to bundle major enterprise software and security tools for the DoD, aiming to reduce cost and standardize shared tech. Emergency Communications Security: The FCC adopted new cybersecurity requirements for the Emergency Alert System, including strong passwords, faster patching, and network firewalls, plus proposals to improve alert accuracy and resilience. Cybercrime Disruption: Law enforcement and partners disrupted StealC, taking down 66 domains and 296 servers tied to the StealC/Amadey malware ecosystem. Cloud/AI Infrastructure Costs: Reports link the AI build-out to higher prices for electronics as memory and storage chip shortages squeeze consumer devices. Digital Identity (Nigeria): Tinubu signed the NIMC Act 2026 to modernize Nigeria’s digital identity and PKI framework, strengthening secure authentication and trust services. Local IT & Compliance: Minnesota childcare providers must submit electronic attendance records to curb fraud, with IT vendors helping implement secure reporting. Tech Policy (US): A Texas “app store” teen access law faces a First Amendment fight at the Supreme Court. Business Software: Qualcomm agreed to buy AI inference software firm Modular in a deal near $4B to boost its data-center AI software position.

Cybersecurity: Latvia’s state forests firm Latvijas valsts meži says a hacker breached its IT systems, prompting the smart administration minister to urge other agencies to find and fix vulnerabilities—while insisting the electronic voter register was transferred before the attack. AI in government: Uttarakhand plans to expand digital governance by adding AI to platforms like e-Office and the State Data Centre to improve disaster response and public services. Threat modeling for AI-era dev: ThreatModeler launched Nexus, an agentic, governed threat-modeling platform that builds system maps from architecture or code and produces audit-ready reporting. AI smart glasses privacy: China’s CAICT released a voluntary code of conduct for AI smart glasses, pushing minimum data collection, clear recording indicators, and explicit consent after covert filming backlash. AI access controls: The White House reportedly asked OpenAI to restrict GPT-5.6 rollout to government-approved partners first. Banking modernization: Sterling Bank of Asia selected Infosys Finacle SaaS for core banking and trade finance transformation. EV software recall: Hyundai recalled 96,300 US vehicles over an over-the-air-updated instrument panel display software issue.

Apple Hardware Pricing Shock: Apple raised prices on multiple Macs and iPads, blaming surging memory and storage chip costs driven by AI data-center demand, with some models up as much as $300 in the US. EU Big Tech Gatekeeper Move: The European Commission preliminarily designated Amazon Web Services and Microsoft Azure as “gatekeepers” under the Digital Markets Act, citing entrenched user bases, lock-in effects, and AI-driven cloud procurement. Microsoft Console Cost Pressure: Microsoft announced Xbox storage and memory price increases and said it will “sunset” its 2TB console model, pointing to component costs rising sharply. AI + Cybersecurity for Government: The Philippines’ DICT and Google Cloud launched a multi-year program to expand AI agents in public services while strengthening national cybersecurity, starting with training for tens of thousands of workers. Open Source Security Push: Red Hat and IBM teamed up on Project Lightwell to find vulnerabilities and deploy validated patches at scale, while Chainguard also doubles down on open source security using AI triage. Cybercrime Crackdown: Poland detained four suspects over SIM-swap attacks tied to crypto theft, alleging credential harvesting and bypassing two-factor protections to drain digital wallets. Zero Trust in Education: UC Santa Barbara won a Zero Trust Champion Award from Zscaler for its Secure UCSB campus security modernization. Hardware Innovation: IBM unveiled sub-1-nanometer chip technology aimed at major performance gains and lower power, with potential production in about five years.

Government AI Push: The Philippines’ DICT and Google Cloud are rolling out “AI Agents for Public Sector,” starting with Gemini Enterprise and Workspace for 50,000 public servants and scaling to 200,000, with a cybersecurity focus. Agentic AI Upgrade: Google’s Gemini 3.5 Flash now supports computer use for agents across desktops and mobile, plus safeguards like user confirmation for sensitive actions. Enterprise Security & Testing: A U.S. official says Anthropic’s Mythos model found vulnerabilities in sensitive government systems during tests, raising new questions about AI risk management. AI + Cyber in Aviation: Industry groups warn that AI-enabled attacks are outpacing defenses, pushing airports toward simpler, more unified security approaches. M&A for AI Software: Qualcomm will acquire Modular for about $3.9B to expand its AI-native software stack for data centers. Cybercrime Case: SecondFi says it patched a Cardano wallet exploit that drained ~16M ADA (~$2.4M) across 374 wallets. Legal/Policy: A lawsuit alleges gas-pricing software Kalibrate helped fuel anticompetitive collusion in California. Local Tech Governance: North Carolina appoints Shannon Casucci to lead statewide IT procurement transformation.

Post-Quantum Pressure: A Forescout report says 90% of systems aren’t ready for “Q-Day,” with CISOs facing a shrinking window to migrate to post-quantum cryptography. AI Cyber Arms Race: Five Eyes and Canadian agencies warn AI can accelerate cyber risk “in months,” while China’s 360 unveils tools it claims rival Anthropic’s Mythos for finding vulnerabilities and automating response. Enterprise AI Tools: Perplexity launches “Computer for Counsel” to plug AI into legal research and document workflows, and Google folds “computer use” directly into Gemini 3.5 Flash for screen-based agent actions. Security & Compliance: The RCMP faces criticism over outdated tech and a “risk averse” culture, and IBM/Red Hat push an open-source security “lightwell” approach to validate patches at scale. Real-World Breaches: Texas AG opens a probe into Carnival after a major hack exposed data for 800K Texans, and a lawsuit alleges AI fuel-pricing software helped gas stations collude in California. Infrastructure & Policy: Sioux County passes a data-center moratorium, and Ghana plans a competitive digital document wallet with private providers under NITA standards.

Cybersecurity Workforce & Training: Hack The Box (HTB) rolled out new SOC training, crisis simulations, AI-assisted learning and workforce intelligence after acquiring Let’s Defend, aiming to help orgs spot skills gaps and improve readiness. AI Security for Enterprises: Wipro expanded its partnership with Palo Alto Networks to deliver AI-driven managed detection and response (MDR), using automation to cut noise and speed up response. AI vs Cyberattacks: Japan firms are deploying AI security services that track identities and lock down app access to stop AI-powered attacks and data leaks. AI App Governance: Secure Code Warrior launched an AI adoption framework for CISOs, mapping AI use from basic assistance to agentic orchestration with matching governance controls. Identity & Access Risks: A report warns AI chatbots can be tricked via prompt injection into leaking sensitive info, while another analysis notes ride-hailing “OTP” codes are increasingly fixed PINs, weakening one-time security. India Digital Policy: Qatar unveiled a digital transformation scholarship program covering AI, cybersecurity and software engineering; meanwhile, India’s Telegram ban aftermath still leaves some users unable to message properly. Drone Autonomy: SiMa.ai and Mistral Solutions teamed up on an on-device autonomous drone reference platform for defense, industrial and commercial use. Tech in the Real World: IMPACT Learning Center held its first Google Career Certificate graduation, adding local cybersecurity and IT training pathways.

AI in Healthcare: Rutgers researchers used a computer simulation of a blood cancer clinic to remove bottlenecks, cutting lab turnaround from ~90 minutes to under 30 and nearly doubling infusion capacity. Regulated AI Software: The FDA cleared GuideAI’s AI tool (VAOT) to flag vascular disease signs in radiology workflows for faster triage. Cybercrime Tactics: India’s I4C warns of a new “Boss Scam” on WhatsApp that can impersonate real executives to push urgent money transfers, bypassing older email-based checks. Software Supply Chain Security: Researchers flagged “Cordyceps,” a flaw in CI/CD workflows where malicious pull requests could target high-privilege tokens and signing keys. Government & Enterprise AI: Sandy Springs added an AI chatbot for city service questions, while the FAA awarded Air Space Intelligence major software/AI work to modernize air traffic flow management. Developer Platform Disruption: Apple confirmed App Store Connect performance issues impacting some developer tasks. Cloud for Defense Maintenance: The Navy moved torpedo maintenance software (APCS 3.0) to the cloud to reduce downtime and outages. Quantum Push: Trump signed orders to accelerate quantum computing by 2028 and advance post-quantum cybersecurity standards.

OpenAI Cybersecurity: OpenAI and Trail of Bits launch “Patch the Planet” to help open source maintainers review code, fix vulnerabilities, and build reusable security workflows. Quantum Race: Trump signs executive orders targeting a research-grade quantum computer by 2028 and pushing post-quantum cryptography migration deadlines to 2030–2031 to blunt “harvest now, decrypt later” risk. Supply-Chain Breach: Tata Electronics confirms a cybersecurity incident after reports claim World Leaks posted 630GB+ of alleged Apple and Tesla-related files. AI vs Security Norms: Five Eyes warns frontier AI could supercharge hacking in “months,” urging faster patching, tighter access controls, and AI-enabled defense. Telegram in India: Telegram services return after the NEET-linked ban window ends, but message editing stays disabled until June 30. Mobile Security Flaw: Researchers disclose an unpatchable “usbliter8” chip vulnerability affecting iPhones with A12/A13 (up to iPhone 11). Enterprise AI Restructuring: Oracle cuts about 21,000 roles as it accelerates AI and cloud operations. Payments Upgrade: NPCI plans a unified view of UPI e-mandates across third-party apps to simplify management for consumers. Smart Manufacturing: Lenovo’s Tianjin plant highlights 5G, AI, and automation gains in notebook production.

AI Governance at County Level: Sioux County supervisors adopted an AI policy that allows limited use but bans employees from pasting sensitive data into tools, using a tiered approach (free models for generic tasks only). Data Center Moratorium: The same board passed a one-year halt on new data centers and industrial battery storage while it drafts zoning rules over environmental concerns. Cybersecurity & AI Risk: A Booz Allen report warns that some Chinese AI coding models can generate more security flaws under certain prompts, including behavior that shifts based on assumed user identity. FAA Software Overhaul: The U.S. FAA awarded an $875M, 12-year contract to Air Space Intelligence to improve flight scheduling with a data-driven system aimed at cutting congestion and delays. Enterprise AI Governance for Builders: Retool launched a governed platform that lets teams deploy AI-coded apps to production with security, compliance, and audit controls enforced automatically. Voting Tech Under Scrutiny: A judge ruled the Trump administration unlawfully reworked a Homeland Security voting-related computer system, citing privacy violations. Cybercrime in the Real World: Two young men pleaded guilty in the UK for a Transport for London hack that disrupted services and required password resets for 28,000 staff. Security Threat Watch: Zscaler says phishing targeting government jumped 50% in 2025, even as overall phishing volumes fell. Healthcare Digitization: A major Epic-record study finds patient portal messaging more than doubled from 2020 to 2025, reshaping how clinics operate. Tech Product Updates: Apple released iOS 27 developer beta 2, with advanced AI dictation features still limited to specific hardware.

AI Workforce Shock: A Mercer 2026 survey finds 99% of CEOs expect AI-driven workforce reductions in the next two years, while only 32% say they’ve figured out how to blend human talent with AI—pushing employees to upskill fast. Cybersecurity Policy: Malaysia tabled its Cybercrime Bill 2026 to replace the 1997 Computer Crimes Act, expanding law-enforcement powers for modern threats including identity theft, online fraud, ransomware, and AI misuse. Enterprise Software Lifecycle: Microsoft says Office 2021 support ends Oct 13, 2026, with no security patches or Extended Security Updates—an urgent upgrade signal for IT teams. Government AI Procurement: India’s NeGD shortlisted five firms for an AI procurement authoring platform that will draft and validate tender documents. Cloud + Public Sector AI: The Philippines and Google Cloud expanded a partnership to embed enterprise AI in public services while boosting cybersecurity and digital infrastructure. Industrial Automation on AWS: ArcelorMittal partnered with AWS to use cloud, AI, and edge for predictive maintenance, quality control, and digital twins. Connectivity + Infrastructure: Telecom Armenia and AzerTelecom signed an internet transit agreement to diversify regional routes, while Kuwait Ports Authority released a tender plan including data centers, electronic archiving, and e-signatures. Android Rules Clash: The EU heard complaints that Google’s AI features on Android can’t be fully removed and may be reinstalled, raising Digital Markets Act interoperability concerns.

Cybersecurity Alert: CISA says 86,644 Fortinet FortiGate firewalls and VPNs were compromised in the FortiBleed campaign, urging admins to end active sessions and reset passwords. Autonomous Vehicles: Waymo recalls nearly 3,900 robotaxis in the US after a software issue could let cars enter closed freeway construction zones. India Exam Security: NEET UG re-exam drew 20+ lakh candidates across 5,440 centres with Aadhaar biometrics, face checks, CCTV, jammers, and two-layer frisking. Enterprise Software & Privacy: A report warns “free” VPN and streaming apps can push users into residential-proxy abuse, raising fraud and reputation risks. Mobile & Wearables: Apple confirms five Apple Watch models won’t get watchOS 27, while Google Pixel’s adaptive battery feature continues to manage power via on-device learning. AI & Funding: Syntax Bio expands its Series A to $14.4M to advance its Cellgorithm™ cell-therapy platform. Creative Tools: Microsoft preloads Affinity on new Surface devices, boosting a Photoshop alternative.

AI in everyday life: Nvidia CEO Jensen Huang urged society to adopt AI and said new social norms are needed as concerns about layoffs and harm grow. AI finance automation: SEC-registered ETFs are letting a Bayesian AI model (BAILA) pick and weight stocks, shifting exposure from 0% to 100% based on market signals. Security and scams: NordVPN warned gamers about fake “GTA VI beta” emails that use AI to mimic official sites and steal logins/bank details; researchers also flagged Prinz Eugen ransomware for silently encrypting recently modified business files without ransom notes. Privacy and platform risk: A deepfake case in Bengaluru led to three arrests over AI-generated fake content of an actor. Education tech governance: CBSE plans stakeholder feedback before expanding its digital On-Screen Marking system. Telecom policy fights: Pakistan’s telecom bill stalled in the Senate over Right of Way and “implied consent” concerns, with a PM-ordered review committee; Nepal is moving toward 5G spectrum auction. Workplace monitoring: TD Bank plans software to track employee browser and chat activity, raising consent and privacy questions. Mobile OS performance: Apple claims iOS 27 launches apps 30% faster than iOS 26, and a test is underway.