Most exposed secrets were embedded in JavaScript (84%), followed by HTML (8%) and JSON (7%), with AWS credentials comprising over 16% of verified leaks.
Security researchers at Stanford University scanned 10 million webpages and uncovered nearly 2,000 valid API credentials across 10,000 sites, exposing access to critical services including AWS, GitHub, Stripe, and OpenAI.
The findings, detailed in the preprint “Keys on Doormats: Exposed API Credentials on the Web,” warn that leaked keys grant programmatic access—often more dangerous than compromised usernames and passwords—potentially enabling large-scale data exfiltration and even real‑world harm.
Lead author and PhD candidate Nurullah Demir said that attackers could directly access cloud databases and key management systems; one global bank reportedly exposed cloud credentials on its own webpages. In another case, repository keys tied to firmware for drones and remote-controlled devices could have allowed adversaries to push malicious updates.
Most exposed secrets were embedded in JavaScript (84%), followed by HTML (8%) and JSON (7%), with AWS credentials comprising over 16% of verified leaks. While coordinated disclosures cut exposed keys by roughly 50%, researchers found many developers were unaware their credentials were public—and that exposures typically persist for about 12 months, sometimes years.
Why it matters:
API keys often bypass UI safeguards, offering direct, automated access to sensitive resources.
Leaks can cascade: from cloud takeover and data theft to supply-chain attacks via poisoned firmware or code.
What teams should do now:
Remove secrets from client-side code; use server-side proxies and short-lived tokens.
Enforce least privilege and key rotation; monitor usage anomalies.
Add CI/CD secret scanning, SAST/DAST, and CSP to block rogue script sources.
Implement incident playbooks for key revocation and attribution.
The researchers’ message is blunt: treat API keys like crown jewels—and assume the web will find anything left in plain sight.
The productivity challenge in industrial operations has never been about workforce capacity—it has been about the inability to continuously align effort with execution at scale.
AI introduces an intelligent layer of visibility that captures how time, movement, and behaviour translate into output.
Consider what happens when a large construction site in the Middle East begins to mobilise. Thousands of workers pass through entry gates, supervisors scan rosters, and operations appear to be running at full capacity.
By mid-morning, however, subtle inefficiencies begin to surface—workers waiting for permits, teams misaligned with tasks, delayed shift starts that go unrecorded. On paper, productivity is intact. In reality, it is already slipping.
This gap between recorded activity and actual output is one of the least visible challenges in industrial operations today. And it is precisely where AI is beginning to play a defining role.
The Middle East’s industrial expansion has been defined by scale—megaprojects under NEOM like The Line and Oxagon, accelerated infrastructure development under Saudi Vision 2030, and rapidly growing logistics and manufacturing ecosystems. Yet beneath this visible progress lies a quieter, more complex issue – a structural productivity gap that traditional systems have been unable to quantify or correct.
Across the region, AI-powered systems are transforming workforce monitoring from static attendance tracking into continuous operational intelligence—capturing not just who is present, but how work unfolds in real time.
Limits of traditional workforce oversight
Industrial workforce management has over the time relied on periodic supervision such as attendance logs, manual reporting, and physical site inspections. These methods were sufficient in less complex environments, but they are increasingly misaligned with the scale and diversity of modern Middle Eastern operations.
Gary Ng, CEO of viAct.
A single site today may involve multiple contractors, thousands of workers, and overlapping shifts across vast physical areas. In such conditions, supervision becomes intermittent, and productivity becomes an inferred metric rather than a measured one.
The result here, is a persistent disconnect in the ecosystem. Workers may be present but not effectively deployed. Teams may be assigned but not synchronised. Delays occur not as isolated incidents, but as recurring patterns that remain largely invisible within traditional systems.
McKinsey & Company has highlighted how large-scale industrial projects routinely experience significant productivity losses due to fragmented workflows, poor visibility, and inconsistent execution on the ground. Most of the projects than run over their budget by 70 per cent and over schedule by 60 per cent.
This ascertains the suggestion by the International Labour Organisation that a sustainable productivity environment through integrated multilevel interventions is important across different sectors to address these issues.
AI-enabled modules specialising in industrial productivity monitoring introduce a layer of continuous intelligence that connects identity, location, and activity. Contactless face recognition ensures accurate attendance while eliminating proxy check-ins and manual errors. More importantly, it establishes a verified digital baseline from which workforce movement and deployment can be analyzed.
From there, the intelligent systems reconcile gate-level attendance with on-ground presence, ensuring that workers are not only on-site but operating within their assigned zones. This capability addresses one of the most overlooked inefficiencies in industrial operations—the assumption that headcount reflects productivity.
By aligning roster data with real-time activity, organisations can detect misallocation, close coverage gaps, and deploy the right skills where they are actually needed.
Using AI to engineer discipline at scale
Productivity is not only a function of workforce size; it is a function of consistency. In high-density industrial environments like Middle East, even minor deviations in shift discipline like recurring late arrivals, not adhering to SOPs, extended breaks and early exits among workforces can accumulate into significant output loss.
Through continuous video analytics, patterns of shift adherence can be observed and benchmarked across teams, contractors, and operational zones using KPIs such as schedule adherence (planned vs actual shift start time), effective working time and relate them to labour utilisation rate.
In a Dairy and Beverage Facility in UAE, operational managers were facing constant challenges of workforce hygiene maintenance. Despite strict protocols, variations in adherence such as missed sanitisation steps, improper PPE usage, and inconsistent zone discipline—were impacting both product quality and audit readiness.
To match the required levels of compliance, the unit deployed AI monitoring. This led to 30 per cent improvement in workforce discipline with an achievement of more than 95 per cent hygiene compliance accuracy. More importantly, hygiene was no longer dependent on manual enforcement—it became a measurable, trackable operational KPI, embedded directly into daily workflows.
This transition marks a critical evolution: discipline is no longer managed through policy alone, but through data-driven operational design.
Uncovering the hidden cost of idle time
Beyond discipline, a substantial portion of productivity loss originates from idle time—moments when workers are present but unable to proceed due to external constraints.
These constraints often stem from systemic inefficiencies like delays in material availability, bottlenecks in approvals, or gaps in coordination between teams. Individually, they may seem insignificant. Collectively, they represent one of the largest drains on productivity.
AI systems can detect these patterns by analysing workforce movement, inactivity, and workflow disruptions. Repeated waiting periods, unnecessary movement across zones, and clustering of inactivity signals can all indicate deeper operational issues.
By surfacing these insights, organisations can move beyond reactive problem-solving toward proactive optimisation—addressing not just worker behaviour, but the structural inefficiencies that shape it.
Accountability in a multi-contractor ecosystem
The Middle East’s industrial landscape is heavily reliant on multi-contractor models, where different vendors operate simultaneously within the same site. This creates inherent challenges in maintaining consistent standards of productivity and accountability.
AI introduces a unifying framework by enabling contractor-level benchmarking based on consistent, objective metrics. Output per man-hour, adherence to schedules, and workforce utilization can be measured across all contractors, regardless of size or scope.
This level of transparency has implications beyond productivity. It strengthens payroll accuracy, supports compliance with wage protection systems, and reduces disputes by providing verifiable records of workforce activity.
A structural shift in how productivity is governed in 2026
What is emerging is not simply a technological enhancement across the industrial sites in Middle East, but a redefinition of industrial productivity itself.
The productivity challenge in industrial operations has never been about workforce capacity—it has been about the inability to continuously align effort with execution at scale. AI introduces an intelligent layer of visibility that captures how time, movement, and behaviour translate into output.
This perspective reflects a broader transition across the industry, where, productivity is no longer assessed retrospectively through reports and audits. It is increasingly governed in real time, through continuous data and adaptive decision-making.
Gary Ng is the CEO and Co-Founder of viAct, one of Asia’s top Sustainability-focused AI company that provides “Scenario-based Vision Intelligence” solutions for risk prone workplaces.
Memory shortages will persist well into 2027, with only some price easing beginning in 2028 and no return to 2025 pricing levels
Global PC shipments are now expected to decline 11.3 per cent in 2026, a steep downgrade from IDC’s prior -2.4 per cent outlook in November 2025, as memory shortages, rising component prices, and broader supply constraints weigh on production well into 2027, according to the IDC Worldwide Quarterly Personal Computing Device Tracker.
Tablet shipments are forecast to fall 7.6 per cent this year.
IDC said the forecast predates the latest escalation in the Middle East conflict, adding further uncertainty for technology supply chains.
“The overall tech industry… continues to face uncontrollable headwinds,” said Ryan Reith, group vice president, Devices and Consumer, citing growing industry and geopolitical risks and “complete uncertainty around when these pressures will subside.”
Despite softer volumes, higher average selling prices are set to lift market value: PCs are projected to grow 1.6 per cent to $274 billion and tablets expanding 3.9 per cent to $66.8 billion in 2026. “The era of bargain-priced PCs and tablets is behind us for now,” said Jitesh Ubrani, research manager for IDC’s Worldwide Mobile Device Trackers, adding that memory shortages will persist well into 2027, with only some price easing beginning in 2028 and no return to 2025 pricing levels.
IDC expects vendors to prioritise supply chain resilience, diversify component sourcing, and consider down-spec’ing to control costs while maintaining affordability, shaping end-user adoption in the years ahead.
Solar PV is now the fastest-growing technology in the UAE, driven by record-low tariffs, high irradiation, ample land, and policy certainty
Gas-fired capacity is forecast to edge from 44.4GW in 2025 to nearly 46GW by 2035, while nuclear capacity remains around 5.3GW at Barakah, providing approximately 34TWh annually.
The United Arab Emirates (UAE) will expand solar photovoltaic capacity from 6.7GW in 2025 to 32.3GW by 2035, a compound annual growth rate above 17 per cent, according to GlobalData. Solar generation is projected to rise from 15.8TWh in 2025 to 75.4TWh by 2035.
Major projects underpin growth, including Al Dhafra Solar PV, billed as the world’s largest single-site facility; the 1.2GW Noor Abu Dhabi Solar Park, estimated to cut about 1 million metric tonness of CO2 annually; and Dubai’s Mohammed bin Rashid Al Maktoum Solar Park, targeting enough output to power nearly 800,000 homes by 2030.
Under the UAE Energy Strategy 2050, the country plans about $54 billion in clean and alternative energy investments. The 2023 update aims to triple renewable capacity to roughly 14GW by 2030, with targets of 50 per cent clean electricity and a 70 per cent reduction in the power sector’s carbon footprint by mid-century.
GlobalData analyst Mohammed Ziauddin said solar PV is now the fastest-growing technology in the UAE, driven by record-low tariffs, high irradiation, ample land, and policy certainty.
Gas and nuclear will continue to anchor reliability. Gas-fired capacity is forecast to edge from 44.4GW in 2025 to nearly 46GW by 2035, while nuclear capacity remains around 5.3GW at Barakah, providing approximately 34TWh annually.
The UAE plans to integrate large-scale solar with grid stability measures, including gas flexibility, nuclear baseload, and storage such as pumped hydro and batteries, Ziauddin said.
Uses screws instead of glue or rivets to secure the battery and keyboard, and simplifying replacements for parts like the camera and fingerprint sensor.
Apple’s new MacBook Neo—the laptop it announced last week with a starting price of $499 for students—is the most repairable Mac notebook since 2014, according to an iFixit analysis released Friday.
iFixit, which publishes repair guides and sells parts and tools, also rates devices for ease of repair; laptop makers including Dell and Lenovo have used those ratings to improve their designs.
In its Friday teardown, iFixit found Apple made key changes from previous MacBooks, such as using screws instead of glue or rivets to secure the battery and keyboard, and simplifying replacements for parts like the camera and fingerprint sensor.
Apple is widely seen as targeting education markets also served by Google’s low-cost Chromebooks. According to iFixit CEO Kyle Wiens, Chromebooks are frequently repaired in schools, with some districts—such as Oakland, California—training student interns to perform fixes.
Even so, the MacBook Neo scored only 6 out of 10 on iFixit’s repairability scale; some recent Lenovo ThinkPads have scored 9s or 10s. Apple’s pursuit of thinner, lighter devices over the past decade has generally made repairs harder.
Wiens noted that the Neo’s 8 GB of DRAM is soldered directly to the logic board as part of the main processor package—consistent with recent Mac designs—preventing memory upgrades. He argued this could hinder the Neo’s ability to run increasingly complex on-device AI applications, despite Apple’s emphasis on the privacy benefits of local AI processing.
He suggested Apple could improve future models by adding an upgradeable layer of memory. “Apple’s future for privacy-centered AI has to be local models,” Wiens said. “I would argue this is a flaw across Apple’s entire Mac product line.”
Adobe has agreed to a $150 million settlement to resolve a US government lawsuit alleging it concealed hefty subscription termination fees and made cancellations difficult, the Department of Justice said.
The accord includes a $75 million civil penalty and $75 million in free services for customers, and requires court approval.
Filed in June 2024 by the DoJ and FTC, the complaint accused Adobe of burying early-termination fees for its “annual paid monthly” plan—sometimes hundreds of dollars—and steering customers through cumbersome online and phone cancellation processes.
Prosecutors said the practices violated the Restore Online Shoppers’ Confidence Act, which requires clear disclosure of material terms and affirmative consumer consent for recurring charges. The settlement also resolves claims against two Adobe executives.
Adobe said it has streamlined and made its sign-up and cancellation flows more transparent, while denying wrongdoing. Subscriptions accounted for 97 per cent of Adobe’s $6.4 billion revenue in the quarter ended February 27.
The settlement was announced a day after CEO Shantanu Narayen said he would step down after more than 18 years, amid investor concerns over how AI could affect Adobe’s outlook.
