Meta plans to replace its human content security employees with AI
Meta is reportedly phasing out its human-led product risk assessments in favour of AI-powered automation, according to internal documents obtained by NPR. This marks a significant shift in how the tech giant – which owns Facebook, Instagram and WhatsApp – evaluates the potential dangers of its product updates and features. For over a decade, Meta has relied on teams of human reviewers to conduct what's known internally as "privacy and integrity reviews". These reviews ensured that new features did not compromise user privacy, promote harmful content or endanger young users. But soon, up to 90 per cent of these reviews will be handled by AI instead of humans, NPR reports.advertisementThis automation will apply to changes in algorithms, new sharing options, and even features related to youth safety and AI ethics. In effect, the same tools Meta uses to build products will now be used to judge their potential risks – with minimal human input.Meta believes the change will speed up product development. Reportedly, developers will receive near-instant AI-generated feedback based on questionnaires they fill out about new products. These AI systems will then flag potential risks and set out requirements to mitigate them, which teams must confirm they've met before launching.
The company insists that humans will still review 'complex or novel cases', and that low-risk decisions are the only ones being automated. It also pointed out that automation allows its human reviewers to focus on more serious or ambiguous content moderation issues.advertisementThis shift comes at a time when Meta is expanding its use of AI across the board. CEO Mark Zuckerberg recently said that within the next 12 to 18 months, most of the code behind Meta's AI efforts, particularly its Llama models, will be written by AI itself. He claimed the company's AI agents are now capable of running tests, spotting bugs and generating better code than the average developer. Zuckerberg also said Meta is building specialised AI agents for internal use, fully integrated into its software development tools. He said these will be tailored to support research and development in AI, rather than general-purpose software engineering.Meta's embrace of AI mirrors a broader trend across the tech industry. Google's Sundar Pichai says AI now writes 30 per cent of the company's code. OpenAI's Sam Altman claims that in some companies, half of all code is AI-generated. And Anthropic CEO Dario Amodei predicts that by the end of 2025, almost all code will be written by AI.Meta says it is auditing AI-made decisions and that its EU operations, which are bound by stricter rules under the Digital Services Act, will retain a more human-led review system. But insiders reportedly suggest that most risk decisions globally are already being handed over to algorithms.As per the report, a current Meta employee said the goal is to "empower product teams" and "simplify decision-making." But they warned the consequences of removing human oversight could be severe: 'We provide the human perspective of how things can go wrong. That's being lost.'
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Time of India
40 minutes ago
- Time of India
Gang exploited victims using artificial intelligence and fake loans
New Delhi: Akshita, Jayshree, Pinki and Dimple led seemingly ordinary lives, but behind the facade, they allegedly harboured secrets that would shock their families and friends. At work, they were telecallers who would ruthlessly target the vulnerable and the elderly by using artificial intelligence, trapping them into a vicious cycle of blackmail. Beyond office hours, the women donned an entirely different persona – a tightrope walk to balance their alleged illicit activities and their everyday lives. Jayshree, a graduate from Alwar University, allegedly worked as a key player in a sextortion gang, using her charm to lure victims into compromising situations, police said. "The racket comprised social media profile analysts, extortion callers and individuals impersonating police officers to intimidate victims with false legal threats unless they complied with extortion demands," said joint commissioner Surender Kumar. These persons allegedly employed a deceitful method to lure and exploit victims. The modus operandi was to check profiles of prospective targets on social media and get them to chat. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like Genera ingresos extras con IA desde casa: una manera astuta de empezar a invertir con éxito [CFD] Digital Group Prueba ahora Undo After establishing a connection, they would suddenly initiate video calls with AI-generated models, and the background voice was of a female. "During interrogation by ACP Anil Sharma, they disclosed that they would initially send friend requests on Facebook and Instagram and obtain the victim's mobile number through Facebook Messenger. Subsequently, they would initiate a video call on WhatsApp, during which they would stream obscene clips, secretly recording the victim's screen and reactions using screen recording tools," said DCP (crime) Aditya Gautam. The recorded content was then used to blackmail the victim under threats of publicly circulating the video on social media platforms, thereby coercing them into transferring money, said police. The sextortion gang, allegedly led by Jahid (20), targeted victims through compromising videos, with Jayshree (24), a former SBI BPO employee, making calls and luring victims into compromising situations. Raheesh (21), Sohil (18) and Jaiveer (24) were also part of this module, cops said. Another modus operandi of the call centre was to trap people by offering fake bank loans. Initially, the telecallers, under directions from Dilshad Ali, contacted potential victims, offering them attractive loan deals. Once they gained the victim's trust, the telecaller would request identification and other personal documents via WhatsApp. The victim was then asked to make a payment through a QR code scanner under the pretext of file processing fees. After the payment was received, the victim's follow-up messages and calls were ignored. The fraudsters would then switch off their mobile devices and change their contact numbers, effectively cutting off all communication. At the heart of the operation were Ujjawal Pandey (30) and Gaurav Barua (24), who allegedly masterminded the racket. Pandey, owner of a cyber café in Krishna Nagar, began selling fake account kits and SIM cards while Barua, a graduate from Ignou, facilitated pre-activated account kit transactions, said police. Their partnership enabled the smooth functioning of the scam, with Yug Sharma (18) allegedly assisting in supplying fake SIM cards and account kits. Ali employed several telecallers, including Saurav (27), Pravesh (28), Akshita (18), Raunak (24), Dimple (20) and Pinki (24), to pose as loan service representatives and dupe customers into divulging sensitive information. Another key player, Amit, owned the call centres and is absconding. He also supervised operations, cops said.
Time of India
3 hours ago
- Time of India
Facebook parent Meta becomes the latest American company to 'go nuclear' for AI
Meta has signed a 20-year agreement to secure nuclear power to meet the growing energy demands for artificial intelligence and other computing needs at Facebook's parent company. According to a report by Associated Press, the deal will also boost the output of Constellation Energy's Clinton Clean Energy Center in Illinois. Announced on June 3, the agreement is part of a broader trend of tech companies partnering with nuclear energy providers as AI usage surges. Financial terms were not disclosed. The Clinton plant, originally set to close in 2017 due to financial losses, was sustained by Illinois' zero-emission credit program, which expires in 2027, as per the report. The Meta deal, effective from June 2027, will increase the plant's clean energy output by 30 megawatts, preserve 1,100 local jobs, and generate $13.5 million in annual tax revenue, the companies said. 'Securing clean, reliable energy is necessary to continue advancing our AI ambitions,' Meta's head of global energy, Urvi Parekh, said. The report notes that tech giants face dual pressures: increasing energy supplies for AI and data centers while meeting ambitious greenhouse gas reduction goals. Constellation, which owns the shuttered Three Mile Island plant, announced in September a plan to restart it to supply Microsoft's data centers, the AP reports. The Pennsylvania plant was the site of the U.S.' worst commercial nuclear accident in 1979. Amazon and Google have also invested in small nuclear reactors, with Amazon announcing plans two days after Google last fall, and Google investing in three advanced nuclear projects with Elementl Power last month, according to the report. The AP report adds that 25 states passed legislation supporting advanced nuclear energy last year, with over 200 nuclear-friendly bills introduced this year, per the Nuclear Energy Institute. However, the U.S. faces challenges in scaling nuclear production, with no next-generation reactors operating commercially and only two new large reactors built in nearly 50 years, completed significantly over budget and behind schedule in Georgia, the AP states. Amazon, Google, and Microsoft are also investing in solar and wind energy, which produce no greenhouse gas emissions, the AP adds. AI Masterclass for Students. Upskill Young Ones Today!– Join Now
Mint
3 hours ago
- Mint
Meta enters 'nuclear' arena? Here's what you should know about the power agreement with Constellation Energy
Meta Platforms on Tuesday said it has struck an agreement with Constellation Energy to keep one of the utility's reactors in Illinois operating for 20 years, in the Big Tech company's first such deal with a nuclear power plant. Big Tech companies are looking to secure electricity as U.S. power demand rises for the first time in two decades on demand from artificial intelligence and data centers. The deal will help keep open Constellation's Clinton Clean Energy Centre which may have been forced to shut after the expiry in 2027 of an Illinois ratepayer-funded zero emissions credit program that awards benefits for generation of power virtually free of carbon emissions. It depends on the U.S. Nuclear Regulatory Commission renewing the plant's license, after Constellation last year applied to renew it through 2047. Meta's power purchase will replace the Illinois support and help with re-licensing and operations. The deal could serve as a model for Big Tech companies to support existing nuclear plants while they also plan to power data centers with new nuclear and other energy sources. Constellation Energy shares rose 5.7% to $330.93 in morning trading. Financial details of the deal were not revealed, but such lengthy power purchase agreements tend to run to multi-billion dollars. The deal also allows Constellation to expand Clinton, which has a capacity of 1,121 megawatts, by 30 MW. The plant powers the equivalent of about 800,000 U.S. homes. Urvi Parekh, head of global energy at Meta, said: "One of the things that we hear very acutely from utilities is they want to have certainty that power plants operating today will continue to operate." Joe Dominguez, CEO of Constellation, said, "We're definitely having conversations with other clients, not just in Illinois, but really across the country, to step in and do what Meta has done, which is essentially give us a backstop so that we could make the investments needed to re-license these assets and keep them operating." Bobby Wendell, an official at a unit of the International Brotherhood of Electrical Workers, said the agreement will deliver a "stable work environment" for workers at the plant. Evercore ISI analyst Nicholas Amicucci viewed the deal as "first of many" following President Donald Trump's recent executive orders aimed to boost U.S. nuclear capacity. Independent power producers like Constellation have announced some of the biggest power deals in U.S. history in the last year, as AI data centers are expected to drive U.S. electricity consumption to record highs in 2025 and 2026. Last year, Constellation announced it had struck an agreement to restart a reactor at the Three Mile Island nuclear plant to serve Microsoft data centers. The reactor, which is being prepped for a restart, requires various approvals. Vistra Corp and NRG Energy have also announced major planned acquisitions in recent months. (Reporting by Timothy Gardner, additional reporting by Laila Kearney and Pooja Menon; Editing by Leslie Adler, Vijay Kishore, Alexandra Hudson)
