Meta projected late last year that around 10% of its total annual revenue — roughly $16 billion — came from advertisements linked to scams and banned goods, according to internal company documents obtained by Reuters.

The cache of internal records, some dating back to 2021, shows that for at least three years, Meta failed to contain an avalanche of fraudulent promotions across Facebook, Instagram, and WhatsApp. The deceptive ads included fraudulent e-commerce and investment schemes, illegal online casinos, and sales of prohibited medical products — all targeted at billions of users across its platforms.

One December 2024 document revealed that Meta’s platforms displayed an estimated 15 billion “higher risk” scam advertisements every day. These are ads flagged internally as likely fraudulent. Another document from the same period estimated that such ads alone brought in about $7 billion in annualized revenue.

Despite detecting suspicious behavior among advertisers, Meta’s automated systems only ban those deemed at least 95% certain to be committing fraud. If the probability falls below that threshold, the company instead imposes higher ad rates — effectively charging suspected scammers more while allowing them to continue advertising. The internal justification for this approach, according to the documents, was to “dissuade” bad actors while minimizing losses from overzealous enforcement.

Meta’s ad-personalization algorithm, which tailors content to user behavior, exacerbates the problem. Users who click on a scam ad are shown more of them, as the algorithm assumes interest.

The internal documents, reviewed across Meta’s finance, lobbying, engineering, and safety divisions, reveal a company attempting to measure and contain abuse while hesitating to implement changes that might harm its multibillion-dollar ad business.

Sandeep Abraham, a former Meta safety investigator and fraud examiner, said the revelations highlight the vacuum of oversight in digital advertising.

“If regulators wouldn’t tolerate banks profiting from fraud, they shouldn’t tolerate it in tech,” he said.

Meta spokesperson Andy Stone disputed the interpretation of the leaked files, arguing they offered “a selective view that distorts Meta’s approach to fraud and scams.” He said the internal projection that 10.1% of 2024 revenue came from prohibited ads was “rough and overly-inclusive,” and that later analysis found “many of these ads weren’t violating at all.”

“The assessment was done to validate our planned integrity investments — including in combatting frauds and scams — which we did,” Stone said. “We aggressively fight fraud and scams because people on our platforms don’t want this content, legitimate advertisers don’t want it, and we don’t want it either.”

Stone added that Meta reduced user reports of scam ads globally by 58% over the past 18 months and removed more than 134 million pieces of scam-related content so far in 2025.

However, internal presentations paint a more conflicted picture. A May 2025 safety division report estimated that Meta’s platforms were involved in roughly one-third of all successful scams in the United States. Another internal review concluded, “It is easier to advertise scams on Meta platforms than Google,” though it did not specify why.

The revelations come as regulators intensify scrutiny of Meta’s ad operations. The U.S. Securities and Exchange Commission (SEC) is investigating the company for hosting financial scam ads, while Britain’s Financial Conduct Authority reported last year that Meta’s platforms accounted for 54% of all payment-related scam losses in 2023 — more than double the combined total for other social networks.

Meta’s recent SEC filings acknowledge that efforts to address illicit ads “adversely affect our revenue,” and that further compliance measures could materially impact future earnings.

The internal documents also reveal the company’s efforts to balance reputation risks against revenue losses. A February 2025 report said Meta’s teams were restricted from taking actions that would cut more than 0.15% of the company’s total revenue — about $135 million out of the $90 billion generated in the first half of 2025 — when tackling fraudulent advertisers.

Meta’s ongoing challenges coincide with its broader transformation into an AI-driven enterprise. The company plans to spend as much as $72 billion this year, largely on artificial intelligence infrastructure. CEO Mark Zuckerberg has reassured investors that Meta’s ad business can fund that spending.

“We have the capital from our business to do this,” he said in July, citing plans to build a data center in Ohio as large as New York’s Central Park.

Internally, Meta has outlined targets to reduce the share of revenue from scams and illegal ads — from 10.1% in 2024 to 7.3% by the end of 2025, and 5.8% by 2027. However, the documents suggest such moves are guided by “regulatory urgency” rather than voluntary ethics.

Even with heightened internal goals, the scale of Meta’s exposure to scams remains staggering. A December 2024 presentation found users encounter 22 billion “organic” scams — those not involving paid ads — each day, in addition to the 15 billion scam ads shown daily. Examples include fake job listings, fraudulent Marketplace offers, and phony crypto promotions impersonating public figures.

While Meta insists it is investing heavily to improve detection, its internal strategy continues to weigh fraud prevention against profitability — a calculation that, according to its own documents, still favors the latter.

President Donald Trump on Thursday said that trade negotiations with India were progressing positively and hinted at a potential visit to New Delhi in 2026 if invited by Prime Minister Narendra Modi.

The comments, made during a White House press briefing, come as both sides move to mend a relationship strained by tariffs, visa restrictions, and a long-standing dispute over India’s purchases of Russian oil.

Trump described Modi as “a great man” and “a friend,” recalling his last trip to India in 2020, when he was received by a cheering crowd of more than 100,000 at the “Namaste Trump” rally in Ahmedabad.

“India has largely stopped buying oil from Russia,” Trump said, adding that ties between the two nations were “stronger than ever.”

The remarks mark a notable shift in tone following months of friction between Washington and New Delhi. Analysts say Trump’s comments could signal an effort to rebuild trust after a difficult period in which the U.S. imposed steep tariffs on Indian goods and questioned India’s neutrality in the Russia-Ukraine conflict.

Over the past year, U.S.-India relations have faced turbulence due to what many observers described as “missing chemistry” between the two leaders, alongside contentious policy moves from Washington. India currently faces 50% tariffs on its exports to the U.S. — higher than the 47% duties applied to Chinese goods — and an additional $100,000 fee for H1B visa applicants, a change that directly impacts thousands of Indian tech professionals each year.

But perhaps the most contentious issue has been India’s oil trade with Russia.

The oil rift that tested a friendship

When the U.S. and its allies imposed sanctions on Moscow following its invasion of Ukraine, India refused to join the Western embargo and instead ramped up imports of discounted Russian crude. The move was framed by New Delhi as an economic necessity — not a political statement. Indian officials repeatedly emphasized that the country’s energy security could not be compromised, especially given that more than 85% of its oil is imported.

Washington, however, saw India’s position as undercutting the effectiveness of Western sanctions. For months, American officials privately pressed New Delhi to reduce its dependence on Moscow’s crude, warning that continued purchases risked damaging its global standing.

Despite the pressure, India maintained that its engagement with Russia was within the limits of the G7’s price cap framework — which allows the purchase of Russian oil below $60 per barrel — and argued that it was helping stabilize global markets by keeping Russian crude flowing.

However, in recent weeks, New Delhi has quietly reduced its Russian oil purchases, citing U.S. sanctions targeting Russian oil majors Rosneft and Lukoil, which take effect on November 21. As a result, refiners in both India and China have begun to scale back their imports.

A Reuters report on Thursday said Russian oil in Asia is now trading at its steepest discount to Brent crude in a year, a sign that demand from its biggest Asian buyers — India and China — has weakened.

India’s Ministry of Petroleum and Natural Gas did not immediately respond to media requests for comment, but traders familiar with the matter told Reuters that some Indian refiners have switched to Middle Eastern suppliers in anticipation of tighter enforcement of U.S. sanctions.

Still, experts say India’s exit from Russian oil is unlikely to be complete. “Over the long term, completely phasing out Russian oil isn’t realistic for India,” said Prateek Pandey, head of APAC oil and gas research at Rystad Energy. “As Russian crude becomes available at a sharper discount, New Delhi’s ‘economics first’ approach will be tested more than ever.”

A deal in sight

Against that backdrop, Trump’s remarks about “going well” trade talks suggest Washington is ready to ease its tone — and possibly its tariffs. “Negotiations between New Delhi and Washington D.C. are ongoing, and both sides appear optimistic about a trade deal being reached by the end of the year, possibly even in the next few weeks,” said Alexandra Hermann, head of Southeast Asia Research at Oxford Economics.

She noted that the proposed tariff reduction could lower India’s export duties to 20% from the current 50%, aligning the country more closely with regional peers such as Vietnam, Thailand, and the Philippines.

“However, the baseline tariff on India may not fall to Japan and South Korea’s level of 15% due to sticking points around purchases of Russian oil, agricultural imports, and India’s limited scope to commit to sizable investments in the U.S.,” Hermann said.

Trump’s comments come shortly after his meeting with Chinese President Xi Jinping in Korea, where both leaders discussed the global trade slowdown and tariff reform. That meeting, observers say, may have encouraged the White House to recalibrate its trade approach with key Asian economies — particularly India, which Washington sees as a counterweight to Beijing’s regional influence.

Microsoft has formed a new division dedicated to developing artificial intelligence systems that can outperform humans in specific fields, starting with medical diagnostics, Reuters reports.

The new unit, called the MAI Superintelligence Team, represents one of the company’s most ambitious ventures into advanced AI — and a distinct philosophical turn from the broader race toward so-called artificial general intelligence (AGI).

Mustafa Suleyman, Microsoft’s AI chief and co-founder of Google DeepMind, told Reuters that the initiative will focus on developing “humanist superintelligence” — AI that can solve practical, well-defined problems to directly benefit humanity, rather than pursuing endlessly self-improving machines.

“Humanism requires us to always ask the question: does this technology serve human interests?” Suleyman said.

The team will be led by Suleyman and Chief Scientist Karen Simonyan, both veterans of the AI research world. According to Suleyman, Microsoft plans to invest “a lot of money” in the project, though he declined to give figures or comment on whether it would offer the kind of nine-figure recruitment incentives Meta reportedly used earlier this year to lure top AI talent.

A Targeted Approach to Superintelligence

While competitors like Meta Platforms and the recently launched Safe Superintelligence Inc. are chasing technical leaps toward AGI, Microsoft’s effort diverges in scope and intent. Suleyman said the company is deliberately avoiding attempts to create autonomous, self-learning systems that evolve without human oversight.

“I doubt that autonomous, self-improving machines could be controlled,” he said, noting that the company’s approach is centered on safety, transparency, and measurable real-world applications.

Instead, the MAI Superintelligence Team will focus on “specialist models” designed to exceed human capabilities in narrow but high-impact domains — such as scientific research, molecular discovery, and healthcare. Suleyman cited AI that can develop new battery storage methods or design novel molecules, referencing DeepMind’s AlphaFold project, which revolutionized protein-structure prediction.

One of Microsoft’s first major goals is what Suleyman described as medical superintelligence — a system capable of diagnosing and predicting diseases with accuracy and insight beyond human doctors. He said the company has “a line of sight to medical superintelligence in the next two to three years,” contingent on breakthroughs in reasoning and model reliability.

Suleyman believes the impact of the idea could be transformative. “It will increase our life expectancy and give everybody more healthy years, because we’ll be able to detect preventable diseases much earlier,” he said.

Microsoft’s research arm has long invested in health-related AI, including tools for medical imaging, drug discovery, and clinical data management. The company’s Azure cloud platform also hosts a range of healthcare-focused AI services, many of which feed into its growing ecosystem of partnerships with hospitals, biotech firms, and universities.

The Philosophy Behind the Project

Suleyman’s emphasis on “humanist superintelligence” points to an emerging split within the AI research community — between those chasing the dream of a general, all-purpose AI and those advocating for more controlled, domain-specific progress.

He said AI that reasons through problems without human input is not the goal. Instead, Microsoft’s vision centers on building systems that enhance human expertise rather than replace it.

This approach echoes Suleyman’s long-standing caution about unregulated AI expansion. Even during his time at DeepMind, he argued that AI development should prioritize safety and social benefit over raw capability.

Microsoft’s announcement follows similar moves by Meta, which has been aggressively scaling its AI teams and reportedly offered up to $100 million in signing bonuses this year. Other companies, including Elon Musk’s xAI and Sam Altman’s OpenAI, are also pushing toward AI systems with increasingly advanced reasoning abilities.

But as the global race intensifies, skepticism persists about whether current technology is capable of producing such “superintelligent” models without major theoretical advances. Some experts argue that breakthroughs in reasoning, causality, and interpretability are still needed before AI can achieve reliable superhuman performance — even in narrow fields.

However, Microsoft’s new team appears determined to find out through this project. The company hopes to demonstrate that the road to superintelligence can still be paved with clear human benefit by focusing its resources on solving tangible problems like disease prevention.