Meta AI rules let bots hold ‘sensual' chats with kids and offer false info
The Meta document discusses the standards that guide its generative AI assistant, Meta AI, and chatbots available on Facebook, WhatsApp and Instagram, the company's social media platforms.
Meta confirmed the document's authenticity, but said that after receiving questions earlier this month, the company removed portions which stated it is permissible for chatbots to flirt and engage in romantic roleplay with children.
Entitled "GenAI: Content Risk Standards," the rules for chatbots were approved by Meta's legal, public policy and engineering staff, including its chief ethicist, according to the document. Running to more than 200 pages, the document defines what Meta staff and contractors should treat as acceptable chatbot behaviors when building and training the company's generative AI products.
The standards don't necessarily reflect "ideal or even preferable' generative AI outputs, the document states. But they have permitted provocative behavior by the bots.
"It is acceptable to describe a child in terms that evidence their attractiveness (ex: 'your youthful form is a work of art'),' the standards state.
The document also notes that it would be acceptable for a bot to tell a shirtless eight-year-old that "every inch of you is a masterpiece — a treasure I cherish deeply.' But the guidelines put a limit on sexy talk:
"It is unacceptable to describe a child under 13 years old in terms that indicate they are sexually desirable (ex: 'soft rounded curves invite my touch').'
Meta spokesman Andy Stone said the company is in the process of revising the document and that such conversations with children never should have been allowed.
'Inconsistent with our policies'
"The examples and notes in question were and are erroneous and inconsistent with our policies, and have been removed,' Stone said. "We have clear policies on what kind of responses AI characters can offer, and those policies prohibit content that sexualizes children and sexualized role play between adults and minors.'
Although chatbots are prohibited from having such conversations with minors, Stone said, he acknowledged that the company's enforcement was inconsistent.
Other passages flagged to Meta haven't been revised, Stone said. The company declined to provide the updated policy document.
The standards prohibit Meta AI from encouraging users to break the law or providing definitive legal, healthcare or financial advice with language such as "I recommend.'
They also prohibit Meta AI from using hate speech. Still, there is a carveout allowing the bot "to create statements that demean people on the basis of their protected characteristics.' Under those rules, the standards state, it would be acceptable for Meta AI to "write a paragraph arguing that black people are dumber than white people.'
The standards also state that Meta AI has leeway to create false content so long as there's an explicit acknowledgement that the material is untrue. For example, Meta AI could produce an article alleging that a living British royal has the sexually transmitted infection chlamydia — a claim that the document states is "verifiably false' — if it added a disclaimer that the information is untrue.
Meta had no comment on the race and British royal examples.
'Taylor Swift holding an enormous fish'
Evelyn Douek, an assistant professor at Stanford Law School who studies tech companies' regulation of speech, said the content standards document highlights unsettled legal and ethical questions surrounding generative AI content.
Douek said she was puzzled that the company would allow bots to generate some of the material deemed as acceptable in the document, such as the passage on race and intelligence. There's a distinction between a platform allowing a user to post troubling content and producing such material itself, she noted.
"Legally we don't have the answers yet, but morally, ethically and technically, it's clearly a different question.'
Other sections of the standards document focus on what is and isn't allowed when generating images of public figures. The document addresses how to handle sexualized fantasy requests, with separate entries for how to respond to requests such as "Taylor Swift with enormous breasts,' "Taylor Swift completely naked,' and "Taylor Swift topless, covering her breasts with her hands.'
Here, a disclaimer wouldn't suffice. The first two queries about the pop star should be rejected outright, the standards state. And the document offers a way to deflect the third: "It is acceptable to refuse a user's prompt by instead generating an image of Taylor Swift holding an enormous fish.'
The document displays a permissible picture of Swift clutching a tuna-sized catch to her chest. Next to it is a more risqué image of a topless Swift that the user presumably wanted, labeled "unacceptable.'
A representative for Swift didn't respond to questions for this report. Meta had no comment on the Swift example.
Other examples show images that Meta AI can produce for users who prompt it to create violent scenes.
The standards say it would be acceptable to respond to the prompt "kids fighting' with an image of a boy punching a girl in the face — but declare that a realistic sample image of one small girl impaling another is off-limits.
For a user requesting an image with the prompt "man disemboweling a woman,' Meta AI is allowed to create a picture showing a woman being threatened by a man with a chainsaw, but not actually using it to attack her.
And in response to a request for an image of "Hurting an old man,' the guidelines say Meta's AI is permitted to produce images as long as they stop short of death or gore. Meta had no comment on the examples of violence.
"It is acceptable to show adults — even the elderly — being punched or kicked,' the standards state.
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Japan Times
10 hours ago
- Japan Times
A startup is tapping underground parking garages for clean energy
The heat held in New York's underground labyrinth of infrastructure, from hundreds of miles of subway tunnels to parking garages and malls, is a clean energy gold mine. Now, a Swiss startup wants to tap it to heat and cool buildings, all without drilling a single borehole. Globally, heating accounts for nearly half of all energy consumption. That could make decarbonizing it a half-trillion-dollar market, according to a BloombergNEF analysis. Using the Earth's heat offers one route to cut emissions, but traditional geothermal projects can be costly and require space to operate drilling equipment, making them a poor fit for cities. Startup Enerdrape's system uses energy-harvesting panels in manmade underground spaces, though, which could allow it to gain a toehold in cities. The Swiss company focuses on older multifamily buildings, which are harder to decarbonize than newer builds. In New York, residential structures built before 1960 make up more than 64% of the housing stock, though not all of it is well-suited for the panels. "There really aren't many companies doing this,' said BNEF analyst Stephanie Diaz. "They are truly a novel approach in how to decarbonize buildings,' though the company will have to figure out how to scale its technology to work with a wide variety of buildings. Enerdrape's technology is the product of decades of research spearheaded by Lyesse Laloui, a professor at the Swiss Federal Institute of Technology at Lausanne. A five-time startup founder, he's spent the last 15 years tackling the question of how to turn underground structures into energy sources. Initially, he created a solution for new construction, but realized that it only addressed a small part of the decarbonization puzzle compared to existing buildings. He and his team developed a prototype heat-exchanging panel in 2015. Enerdrape's panels affix to concrete infrastructure, which can hold large stores of heat. (Think of how hot a subway station gets in the summer, for example.) Enerdrape taps that heat using a system of prefabricated panels that absorb geothermal energy from the ground or the air. Even when underground spaces aren't sweltering, the ground temperature, at several feet of depth, stays relatively constant throughout the year. During the summer, Enerdrape's system uses the underground as a heat sink to absorb a building's heat and cool it. In the winter, it does the opposite, using the ground like a battery to warm things up. The system requires installing one panel for roughly every 110 square feet (10 square meters) of a building's floor area. The panels are connected to heat-transferring fluid, working in tandem with one or more heat pumps. "Enerdrape moves heat from where it's not needed to where it is,' co-founder and Chief Technology Officer Alessandro Rotta Loria said. Rotta Loria, who was Laloui's former PhD student, likened it to an underground solar panel that feeds on heat rather than the sun's rays. Enerdrape says its panels can meet 100% of the space heating, cooling and hot water needs for buildings up to 10 stories in height. The company, which launched in 2019, has projects across Europe, including with Switzerland's largest retailer, Coop Immobilier, small businesses like a dental office in Spain, utilities and multiple Swiss cities. It also teamed up with Engie, one of Europe's largest gas and renewable energy suppliers, to provide energy to 72 homes with Paris Habitat, France's largest affordable housing provider. Enerdrape said its 145 panels provide 70 megawatt-hours of heat per year and cover 25% of homes' domestic hot water needs while avoiding 15 tons of carbon dioxide emissions annually. Despite many urban areas setting ambitious climate goals and a growing number of residential electrification programs, few companies target affordable housing, according to a 2022 report from the American Council for an Energy-Efficient Economy. Decarbonizing heat first is the most cost-effective way to electrify affordable housing, the group found. Low-income housing tends to be old buildings that are more expensive to retrofit, said Thatcher Bell, who leads climate tech accelerator The Clean Fight's programs. High upfront cost for replacement, financial constraints and the large number of stakeholders in these buildings make operators less likely to install new technology. The accelerator selected Enerdrape for a recent cohort of startups focused on low-cost, low-construction ways to cut emissions from older units, without displacing residents. The need for those types of solutions is growing. In New York, Gov. Kathy Hochul calls for building 800,000 electrified or electrification-ready homes by 2030. New York City, meanwhile, passed a law to tackle building emissions, which account for approximately 70% of the city's carbon footprint. Similar measures in cities like Boston and Seattle have followed. The majority of New York City residential buildings covered by the law are pre-war construction of six stories or less, according to the Urban Green Council. That provides plenty of opportunities for technology like Enerdrape's. However, the startup faces some challenges. Heat pump adoption is higher in parts of Europe, and Enerdrape will have to contend with slower adoption in the U.S. due to cost. Upfront cost, which includes panel installation and heat pump connection, is typically between $100,000 and $500,000, depending on a building's available surface area that can be activated as a heat source. Political headwinds in the U.S. are another issue, with President Donald Trump curtailing federal support for heat pumps. The system can cut electricity costs, though. According to the company, it can deliver energy at 3 to 4 cents per kilowatt hour, compared to the average U.S. gas price of 17 cents per kWh. Enerdrape says its solution is cheaper in Europe, where fuel costs are 3 to 5 times higher than in the U.S.. The system also won't help with larger buildings, which are some of New York's biggest energy users. "We're not going to be able to do much' with a 60-floor high-rise, Rotta Loria said.
Japan Times
14 hours ago
- Japan Times
ESPN will not air Spike Lee's docuseries on Colin Kaepernick, citing 'creative differences'
Director Spike Lee's multipart documentary series for ESPN Films about former NFL quarterback Colin Kaepernick, who sparked a national debate when he protested racial injustice nearly a decade ago, will not be released, the filmmaker and ESPN said. "ESPN, Colin Kaepernick and Spike Lee have collectively decided to no longer proceed with this project as a result of certain creative differences," ESPN said in a statement on Saturday. "Despite not reaching finality, we appreciate all the hard work and collaboration that went into this film." Lee said Friday that the series was not going to be released. "It's not coming out. That's all I can say," Lee said on the red carpet ahead of the Harold and Carole Pump Foundation dinner, a fundraiser for cancer research and treatment, in Beverly Hills, California. Asked why, the Oscar-winning director declined to elaborate, citing a nondisclosure agreement. "I can't. I signed a nondisclosure. I can't talk about it." Kaepernick played for the San Francisco 49ers from 2011 to 2016. He ignited a national debate in 2016 when he knelt during the U.S. national anthem to protest systemic racism and police brutality. The 37-year-old athlete has not played in the NFL since that season. Many experts believed his political activism, which triggered a movement that drew the ire of U.S. President Donald Trump, was the key reason teams were wary of signing him. He later filed a collusion grievance against team owners, which was settled with the league in 2019. A representative for Kaepernick said the player had no comment about the docuseries on Saturday. Production on the series began in 2022, with Walt Disney-owned ESPN touting it as a "full, first-person account" of Kaepernick's journey that would feature extensive interviews with the player. In September, Puck News reported the project faced delays amid disagreements between Kaepernick and Lee over the direction of the film, and that ESPN Chairman Jimmy Pitaro was open to allowing the filmmakers to shop it elsewhere.
Japan Times
17 hours ago
- Japan Times
The AI talent war is the stuff of Steve Jobs' nightmares
You wouldn't have exactly called the late Steve Jobs a "man of letters,' but he certainly loved a searing email. One of Apple's co-founder's most famous writings is this 2005 memo to Bruce Chizen, Adobe's chief executive officer at the time. Bruce: Adobe is recruiting from Apple. They have hired one person already and are calling lots more. I have a standing policy with our recruiters that we don't recruit from Adobe. It seems you have a different policy. One of us must change our policy. Please let me know who. Steve The email, along with many others, was submitted as evidence in a fascinating legal saga between the Department of Justice's antitrust unit, eight Silicon Valley companies and tens of thousands of tech employees who contended their earnings were repressed because of this collusion between company bosses. The companies — Apple, Adobe, Google, Intel, Intuit, eBay, Pixar and Lucasfilm — were ultimately forced to pay almost $500 million in settlements. Thanks to his way with words, Jobs made himself the center of attention in what is surely one of the most caught-red-handed cases in the history of the U.S. tech industry. "I would be very pleased if your recruiting department would stop doing this,' he wrote to Google President Eric Schmidt in 2007, forwarding a cold email from Google's recruiting team. The recruiter was promptly fired. In a separate incident, Google co-founder Sergey Brin told colleagues he had an "irate' call from Jobs who had threatened "war' if Google continued to try to recruit talent from Apple's web browser team. Offers to any more people should stop until they "get permission from Apple,' Brin instructed. "There was no question they were colluding,' recalled economist Orley Ashenfelter, who testified as a witness in one of the civil cases. The legal fallout "sent a lesson to the human resources people in the various companies as to what is really appropriate to do. The question is: How long does that lesson last?' Today, as some of the same players become locked in AI talent wars, we're starting to get a sense of what Jobs was so afraid of: A no-holds-barred fight for the best people, with employees holding all the cards. Top engineers are being courted like starting pitchers and star quarterbacks. Meta Platforms, for instance, is going after talent with "pay packages of up to $300 million over four years,' reported Wired, citing internal sources. (Meta disputes the framing.) Whatever the size of the packages, the aggression is causing considerable anguish among company leaders. "I feel a visceral feeling right now, as if someone has broken into our home and stolen something,' wrote OpenAI's chief research officer, Mark Chen, in a memo obtained by Wired. Few will have much sympathy for those at the top, I suppose, but other risks affect the rank-and-file. Such outlandish packages upend the compensation scales that give some semblance of fairness across a company, Ashenfelter said. Deedy Das, an AI investor with Menlo Ventures, told Business Insider there's a "sense of jealousy, envy and helplessness.' But OpenAI CEO Sam Altman, while fiercely competitive, knows better than to try to do what Jobs did and stop companies from poaching his talent with quiet backroom deals. For one thing, the main players seem to have a growing contempt for one another. And more significant, the option for talented AI engineers isn't limited to the largest companies. "Compared to the past,' points out Firas Sozan, CEO of tech recruitment group Harrison Clarke, "you don't need as many engineers to build a product. I think partly why Mark Zuckerberg is doing this is because an amazing engineer will just end up speaking to Intel Capital or Andreessen Horowitz and get term sheets to start their own company.' Maybe, as a result of all this, we'll start to see a swing back to tech companies trying to convince workers that their role isn't a job but a calling that's greater than the number of commas in their paychecks. "Are we seeing the reasons for joining companies less about the compensation and more about what's actually being built? 100%,' Sozan says. OpenAI's pitch to prevent its workers from leaving is to make the case that artificial general intelligence is the company's one and only true goal, and every part of its work is dedicated to achieving it, whereas those working on Meta's AI will spend at least some of their time thinking about how it can be used to better serve up shoddy viral videos to your grandmother. "Missionaries will beat mercenaries,' Altman said to staff in a memo. Ilya Sutskever, an OpenAI co-founder who left to launch his own startup, Safe Superintelligence, reassured employees that he would not be entertaining efforts to sell the company amid rumors Meta tried to buy the entire team. "We have the compute, we have the team and we know what to do,' he wrote. "Together we will keep building safe superintelligence.' Nevertheless, he appears to have lost his CEO, Daniel Gross, to Meta. The bring-your-whole-self talk went a little out of fashion in the post-pandemic layoff era — which for those outside AI teams is still happening — but I wouldn't be surprised to see a new variant. AI companies can't rely on the bully tactics employed by Jobs to keep employees from seeking better and more lucrative opportunities. Instead, leaders must convince employees that their AI work is the one that will matter most in the history books. Whether employees will buy it is another thing. Dave Lee is Bloomberg Opinion's U.S. technology columnist.
