President Donald Trump is signaling a possible wind-down of U.S. military operations against Iran, even as the conflict shows signs of entrenching, with Tehran continuing to absorb and respond to sustained strikes from U.S. and Israeli forces.

In a post on Truth Social, Trump said: “We are getting very close to meeting our objectives as we consider winding down our great Military efforts in the Middle East with respect to the Terrorist Regime of Iran.”

He added that the burden of securing the Strait of Hormuz should fall on other nations. “The Hormuz Strait will have to be guarded and policed, as necessary, by other Nations who use it — The United States does not! If asked, we will help these Countries in their Hormuz efforts, but it shouldn’t be necessary once Iran’s threat is eradicated.”

Yet events on the ground point in a different direction. Iranian media reported that the Shahid Ahmadi-Roshan nuclear enrichment facility in Natanz was struck on Saturday. Officials said there were no radioactive leaks and no immediate danger to nearby residents, but the attack underlines the widening scope of targets, now extending deeper into Iran’s nuclear infrastructure.

From the outset, Trump and his advisers had expected a rapid collapse in Iranian resistance, banking on overwhelming military pressure to force Tehran into submission or negotiations within days. That assumption is proving misplaced. Iran has neither capitulated nor signaled willingness to concede. Instead, it has escalated retaliation, deploying missiles and drones across the Gulf and targeting energy assets central to the global supply chain.

The consequence is a conflict with no clear endpoint. Trump’s assertion that the war could soon wind down sits uneasily alongside the reality of a campaign that continues to expand in both geography and impact. Backing away without securing a decisive outcome risks being seen, politically and militarily, as a defeat—raising the likelihood that operations will persist even as the economic costs mount.

“He’s finding it difficult to drive the news cycle, as he’s accustomed to, because he still can’t explain why he’s taken this country to war and what comes next,” said Brett Bruen, a former foreign policy adviser in the Obama administration who now heads the ?Situation Room strategic consultancy in Washington. “He seems to have lost his mojo on messaging.”

Inside the administration, that tension is increasingly visible. Trump declared in recent days that the war “was Militarily WON,” a claim that contrasts with ongoing Iranian strikes and the near-disruption of maritime traffic through the Strait of Hormuz.

A White House official defended the campaign, saying: “This has been an undisputed military success,” citing the killing of senior Iranian figures, the destruction of much of its navy, and damage to its missile arsenal.

Even so, Tehran has continued to impose costs. Since the war began on February 28, more than 2,000 people have been killed in Iran, according to reports. Iranian forces have leveraged remaining capabilities to strike oil and gas facilities across the region, contributing to a roughly 50% surge in global oil prices. The inflationary effect is already being felt, feeding into higher fuel and energy costs for consumers and businesses worldwide.

Iran’s pressure campaign has also focused on maritime chokepoints. The Strait of Hormuz, through which about a fifth of global oil supply passes, has been partially disrupted by attacks on commercial vessels and the laying of mines. Trump’s insistence that other countries should take over its security has met resistance, particularly from NATO allies who were not consulted before the war began.

Privately, U.S. officials acknowledge frustration within the White House over the lack of allied support. Trump has publicly accused NATO partners of cowardice for declining to deploy naval forces to secure the waterway. The dispute has exposed strains in long-standing alliances at a moment when coordination would typically be expected.

The conflict has also revealed fissures with Israel. Trump said he had no advance knowledge of an Israeli strike on Iran’s South Pars gas field, while Israeli officials indicated the operation had been coordinated. The development has added to uncertainty over how closely aligned the two countries are as the war unfolds.

On the battlefield, the confrontation continues to widen. Iranian gas flows to Iraq were briefly halted after the South Pars strike before resuming, highlighting the vulnerability of regional energy networks. Attacks on infrastructure in Iran and neighboring Gulf states have compounded supply disruptions, tightening markets already on edge.

Within Washington, debate is intensifying over how to proceed. Some advisers are urging the president to find an “off-ramp” and define limits to the campaign. Others argue that stepping back now would embolden Iran and undermine U.S. credibility. Analysts say the administration is grappling with the consequences of early assumptions about how the conflict would unfold.

“They failed to think through the contingencies around ways in which a conflict with Iran could go sideways, where it might not go according to the plan as they laid out,” said John Bass, a former U.S. ambassador.

Aaron David Miller, a veteran Middle East negotiator, offered a sharper assessment: “Trump has built himself a box called the Iran war, and he can’t figure out how to get out of it.”

The war is also beginning to test Trump’s political standing at home. Rising energy costs are feeding voter anxiety, particularly as the administration heads toward elections that could shift control of Congress. Trump had campaigned on avoiding prolonged foreign conflicts, but the current trajectory suggests a campaign that may endure longer than anticipated.

“As the economics play themselves out, people will start to say: ‘Why am I paying high gas prices again? … Why is the Strait of Hormuz now determining whether or not I can take a vacation next month?’” Republican strategist Dave Wilson noted, pointing to the economic pressure building on voters.

Beyond the United States, governments are preparing for further fallout. Keir Starmer is expected to convene senior officials and the governor of the Bank of England to examine support measures for households facing rising energy and borrowing costs. The UK has already announced a £53 million package aimed at helping vulnerable households cope with higher heating bills.

The central contradiction, currently, remains unresolved. Trump is signaling a desire to wind down a war he expected to end quickly. Iran, far from yielding, is sustaining resistance and extending the fight into domains that carry global consequences. With neither side prepared to concede and the cost of disengagement rising, the conflict appears set to drag on.

FedEx Corp., one of the world’s largest logistics companies with roughly 440,000 employees globally, is rolling out an enterprise-wide AI literacy program designed to make its workforce more knowledgeable, efficient, and better positioned for career advancement in an era increasingly shaped by artificial intelligence.

Launched in early December 2025 in partnership with Accenture, the initiative provides personalized, role-based training that evolves with the technology. Vishal Talwar, executive vice president and chief data and information officer at FedEx, who also leads the company’s data logistics solution Dataworks, described the program as a living curriculum refreshed monthly and quarterly.

“This is a living curriculum that will continue to refresh itself every month, every quarter, and we have that in our engagement with Accenture,” Talwar said in an interview. “It was one of the key attributes that we asked for to make sure we designed for something that remains future-relevant.”

The training is delivered through Accenture’s LearnVantage platform and includes interactive live sessions that employees can complete during work hours, back-office time, or at their convenience. FedEx has deliberately remained flexible to determine what delivery methods work best across its diverse workforce — from drivers and package handlers to customs clearance specialists and corporate staff.

Beyond individual learning paths, the program encourages employees to form communities of practice and participate in hackathons. Data scientists across the company recently launched their own community to collaboratively ideate on use cases. Hackathons, a familiar industry practice, allow teams to compete and discover new technological applications.

A distinctive feature of the initiative is the full C-suite commitment from the outset. Every executive took two days off to travel to Silicon Valley for an intensive “speed dating” process with potential partners, ensuring the best fit for FedEx’s needs.

“I have never seen an organization’s full C-suite take off for a two-day to just learn,” Talwar said. “That humility that we have to learn, you can’t build it with just launching a program in isolation. So I truly mean it when I say the whole organization is having a joint experience.”

Early Signs of Impact

Although the program is still in its early stages, Talwar indicates there are already tangible effects. Frontline workers are applying for corporate roles at a higher rate, seeking advancement as they gain AI-related skills. FedEx tracks progress through a metric it calls AIQ (AI quotient), but Talwar stressed the company is measuring participation and learning rather than over-attributing business outcomes to AI alone.

“We are measuring progress around AI, not necessarily just success, because it’s going to be very difficult to say this success is only attributed to AI,” he said. “AI, in my view, needs to be seamlessly embedded in everything that we do.”

The initiative arrives at a challenging time for the logistics industry. FedEx faces ongoing cost pressures, recent plant closures, and layoffs in locations ranging from Kansas to France, and competition from UPS, which announced 30,000 layoffs to add to the 48,000 it conducted in 2025. Tariffs and policy changes further complicate operations. Despite these headwinds, FedEx’s recent earnings reports, including the latest this week, have met with investor approval, with shares up close to 50% over the past year.

Industry Parallels and Lessons from History

FedEx is not alone in prioritizing AI education on a large scale. Accenture’s 2026 Pulse of Change report found that only 28% of organizations have embedded continuous AI learning. Taylor Bradley, vice president of talent strategy and success at AI training company Turing, said the greatest barrier to successful AI adoption is the inertia of the status quo.

Bradley drew a historical parallel to Microsoft’s decision in 1990 to include Solitaire with every Windows installation — a simple game intended to teach users how to use a mouse through drag-and-drop mechanics. Turing applies a similar philosophy, engaging teams with creative ways to leverage large language models.

During one HR offsite, the team built a lifecycle management system from scratch in hours using dummy data in a sandbox environment, eventually scaling it into a production-grade talent automation system that saved roughly 2,000 labor hours while still in beta.

Sunita Verma, CTO of AI contract management platform Ironclad and a former leader at Character.AI and Google, recently ran a “20 days of AI learning” campaign at her company to encourage employees to experiment responsibly.

“When people feel empowered to learn, test and apply AI in meaningful ways, it accelerates adoption and leads to better, more responsible outcomes,” Verma said.

Larger enterprises such as DHL Express have advanced AI-powered career marketplaces to help employees identify in-house opportunities and the skills needed to pursue them. Citigroup operates a smaller-scale AI Champions and Accelerators program that relies on tech evangelism to create a ripple effect across its hundreds of thousands of employees.

FedEx’s Long-Term Vision

What sets FedEx apart is the comprehensive, ongoing nature of its program with no defined endpoint. Talwar explained that technology touches every role at the company, from drivers handling pickups and deliveries to customs specialists managing clearance, and AI can amplify performance in all of them.

“Everybody is dealing with technology,” he said. “They deal with technology differently, and each one of those areas can be amplified further with AI. We decided to make sure that we were comprehensive in providing this program and training for everyone, and more importantly, we were meeting the training program at the point on where it’s helpful and contextual for the individual.”

The initiative is seen as a recognition that AI is not a standalone tool but an embedded capability that will reshape logistics operations. By this move, FedEx aims to stay ahead of competitors, adapt to industry constraints, and turn technological disruption into a competitive advantage.

X has integrated features related to identifying and handling AI-generated content, with recent developments pointing to automatic detection capabilities going live. Recent user reports and posts on X indicate that an AI content detection feature is now active. It automatically scans for AI-generated material and displays a warning prompt before a user posts or reposts, rather than relying solely on manual labeling.

This helps alert users in real-time during composition or sharing, aiming to reduce undetected “AI slop” flooding timelines and improve transparency about what’s real versus synthetic. Users have shared screenshots showing pre-post warnings triggered by the platform’s detection.

This builds on earlier 2026 rollouts, like the “Made with AI” voluntary label, where creators could manually tag posts containing AI-generated or manipulated text, images, or videos. X already watermarks content from its own Grok AI and has policies like requiring disclosures for AI videos of armed conflicts with revenue-sharing penalties for non-compliance.

The new automatic detection appears to be a step toward more proactive enforcement, though it’s not yet clear if it auto-applies labels, reduces visibility, or just warns users. This aligns with broader industry and regulatory pressures to combat misinformation from deepfakes and generative AI.

Deepfake detection techniques aim to identify synthetic or manipulated media (images, videos, or audio) generated by AI models like GANs, autoencoders, or diffusion models. These fakes often appear hyper-realistic but contain subtle inconsistencies that detection methods exploit. Techniques have evolved rapidly, with 2025–2026 surveys emphasizing a shift from handcrafted rules to advanced deep learning, while addressing challenges like dataset generalization, video compression, and emerging generators.

Detection broadly categorizes into forensic and artifact-based, biological/physiological, deep learning (spatial, temporal, frequency, hybrid), and multimodal approaches. Performance is typically measured via accuracy, AUC (Area Under Curve), or F1-score on benchmarks like FaceForensics++ (FF++), Celeb-DF, and DFDC.

State-of-the-art models often exceed 95% on known data but drop 10–15% on cross-dataset or compressed real-world scenarios. These analyze low-level visual inconsistencies without heavy training:Blending boundaries, lighting/shadows, textures, or color mismatches.

Methods use edge detectors (Sobel), Local Binary Patterns (LBP), or frequency transforms like Discrete Cosine Transform (DCT) or Discrete Fourier Transform (DFT) to spot manipulation traces. DFT + SVM achieves ~99% accuracy on FF++ for StyleGAN-generated faces. These are lightweight and interpretable but struggle with high-quality modern deepfakes that minimize visible artifacts.link.springer.com

Deepfakes often fail to replicate natural human signals: Eye blinking patterns: Real humans blink ~15–20 times per minute; fakes may show irregular or absent blinks. Remote Photoplethysmography (rPPG): Extracts subtle skin color changes from blood flow/heartbeat via RGB video analysis. Real videos show consistent pulse signals; fakes disrupt them due to poor temporal synchronization.

Head pose, micro-expressions, or iris and heartbeat variations. FakeCatcher or rPPG + FFT methods reach 98–99% accuracy on FF++ by comparing real vs. synthetic pulse waveforms. These dominate modern detection (70%+ of research) by learning hierarchical features automatically.

Semantic temporal analysis (100% on DFDC via emotional continuity); 3D CNNs for volumetric spatio-temporal features. Transforms images/videos (Fourier/Wavelet) to reveal high-frequency artifacts (e.g., GAN upsampling noise or spectral correlations). Wavelet Analysis (FTWA) boost robustness (up to 99%+ on StyleGAN datasets).

Hybrid pipeline diagrams—CNN extracts spatial features per frame, LSTM/Transformers model temporal sequences for final deepfake classification. Multimodal and Advanced MethodsAudio-visual: Detect lip-audio mismatches or prosody inconsistencies. Global attention for context (e.g., DFDT: 99%+ on FF++/Celeb-DF).

Ensembles/XAI: Random Forest ensembles (99.64% on DFDC, ultra-fast inference) or explainable models linking features to specific GANs. Emerging (2025–2026): Large vision-language models, domain-invariant learning, and real-time tools for platforms combating “AI slop.”

Models overfit to training artifacts; performance drops sharply on compressed/low-res videos or unseen generators (e.g., diffusion models). Robustness: Real-world degradations (noise, lighting) reduce efficacy.

Future directions: Hybrid CNN-Transformer architectures, multimodal fusion, lightweight models for deployment, and benchmarks for unknown forgeries. Detection lags behind generation, but ensembles and physiological hybrids show promise.

In practice, tools including platform integrations like X’s AI content warnings combine multiple techniques for best results. No single method is foolproof—human review or metadata (e.g., C2PA) often supplements. Research continues rapidly to keep pace with evolving generative AI.