字号·· | 护眼
亚洲新闻台

专家警告AI威胁加剧:亚洲过去一年诈骗损失估计达1365亿美元Scam losses across Asia estimated at US$136.5b in past year; experts warn of growing AI threat

点「原文对照」整页切到原文,或双击某段只看那段的原文。

一个人手持智能手机,手机通话界面上显示着“诈骗警报”信息。(资料照片:iStock)A person holding a smartphone displaying a "scam alert" message on its call screen. (File photo: iStock)

曼谷:诈骗分子正日益利用人工智能实现欺诈自动化,并大规模锁定受害者,这促使各方呼吁加强国际合作,因为据估计,过去一年亚洲地区因诈骗造成的损失高达1365亿美元。

BANGKOK: Scammers are increasingly using artificial intelligence to automate fraud and target victims on a massive scale, prompting calls for stronger international cooperation as estimated losses from scams across Asia reach US$136.5 billion over the past year.

全球反诈骗联盟(GASA)警告称,政府和科技公司无法单独应对这一日益增长的威胁,因为犯罪集团利用跨境运作和日益复杂的技术逃避侦查。

The Global Anti-Scam Alliance (GASA) warned that governments and technology companies cannot tackle the growing threat alone, as criminal syndicates exploit cross-border operations and increasingly sophisticated technology to evade detection.

“各国无法单独打击此类犯罪。大型科技公司也无法单独打击。诈骗分子总是跨国作案,(因此)他们被抓获的机会几乎为零,”该联盟执行总监约里亚·亚伯拉罕表示。

“Countries cannot fight this alone. Big tech companies cannot fight this alone. Scammers always work across borders (so) their chances of getting caught are near zero,” said the alliance’s managing director Jorij Abraham.

“执法部门……国际刑警组织和欧洲刑警组织虽然做了大量工作,但能力有限。”亚伯拉罕还警告称,柬埔寨等国对大型诈骗园区的打击,只是分散了犯罪活动,而非将其铲除。

“Law enforcement … Interpol and Europol, while they're doing great work, are limited.”Abraham also cautioned that crackdowns on large scam compounds in countries such as Cambodia have dispersed criminal operations rather than eliminated them.

他表示,此后诈骗活动在中东、墨西哥和东欧等地出现,犯罪集团越来越多地以小团伙形式在酒店和租赁房产中运作,而非大型园区。他补充说,这使得它们更难被侦查和瓦解。

He said scam operations have since emerged in the Middle East, Mexico and Eastern Europe, with syndicates increasingly operating in smaller groups out of hotels and rented properties instead of large compounds. This makes them harder to detect and dismantle, he added.

他的警告是在全球反诈骗联盟上月发布的一份报告之后提出的。报告发现,在13个亚洲市场接受调查的成年人中,71%在过去12个月内曾遭遇诈骗。

His warning comes after a GASA report released last month found that 71 per cent of adults surveyed across 13 Asian markets had encountered a scam in the past 12 months.

该非营利组织总部设在荷兰海牙,汇聚公私部门利益相关者共同打击诈骗和欺诈。

Based in The Hague, Netherlands, the non-profit brings together public and private sector stakeholders to combat scams and fraud.

这项研究由全球反诈骗联盟与防诈骗科技公司Gogolook合作开展,调查了包括新加坡、马来西亚、泰国、印度尼西亚、中国和印度在内的市场中的17460名成年人。

The study, conducted in collaboration with scam prevention technology company Gogolook, surveyed 17,460 adults across markets including Singapore, Malaysia, Thailand, Indonesia, China and India.

研究发现,尽管66%的受访者自信能识别诈骗,但遭遇诈骗者中有62%随后与之互动。在与诈骗互动的人群中,26%遭受了金钱损失。

The research found that while 66 per cent of respondents were confident they could recognise scams, 62 per cent of those who encountered one went on to interact with it. Of those who interacted with a scam, 26 per cent lost money.

AI让诈骗更难识别 周二(11月10日),亚伯拉罕在曼谷举行的GASA(全球反诈骗联盟)地区峰会间隙接受中央通讯社采访时警告称,AI的快速发展使欺诈信息愈发逼真,同时让犯罪分子能以前所未有的规模锁定潜在受害者。

AI MAKING SCAMS HARDER TO DETECT Speaking to CNA on Tuesday (Nov 10) on the sidelines of GASA’s regional summit in Bangkok, Abraham warned that rapid advances in AI are making fraudulent messages increasingly convincing, while allowing criminals to target potential victims on an unprecedented scale.

他引用研究显示,AI生成的钓鱼邮件和短信效果比诈骗分子自己撰写的高出约五倍。

He cited research showing that AI-generated phishing emails and text messages are about five times more effective than those written by scammers themselves.

更令人担忧的是,他说,AI被越来越多地用于将整个诈骗流程自动化,从识别潜在受害者、发送个性化信息,到进行对话并诱导目标访问欺诈网站。

More concerning, he said, is the growing use of AI to automate the entire scam process, from identifying potential victims and sending personalised messages to conducting conversations and directing targets to fraudulent websites.

他说:“整个过程……完全自动化,这意味着诈骗不仅难以识别,还能大规模实现个性化定制。”

“The entire process is … completely automated, and that means the scam is not only unrecognisable, but also personalised on a massive scale,” he said.

亚伯拉罕补充说,这些发展意味着公众教育和宣传活动虽仍重要,但已不足以保护人们免受诈骗侵害。

Abraham added that such developments mean public education and awareness campaigns, while important, are no longer sufficient to protect people from falling victim.

“我非常支持教育,它依然必不可少,但你不能指望普通人识别出诈骗,”他说。

“I’m a big fan of education and it remains essential, but you cannot expect normal people to recognise scams,” he said.

“诈骗分子手段太高明了。(他们)利用AI让诈骗与真实情况难以区分。我们更需要在基础设施层面保护民众。” 电话仍是最常见诈骗渠道 尽管网络诈骗手段日益复杂,传统电话诈骗仍是亚洲地区诈骗分子接触潜在受害者最常用的方式。

“Scammers are getting too good. (They’re) using AI to make (scams) indistinguishable from the real thing. We need to protect people on an infrastructural level way more.”PHONE CALLS MOST COMMON SCAM CHANNEL Despite the growing sophistication of online fraud, traditional phone calls remain the most common method used by scammers to reach potential victims across Asia.

根据报告,56%的受访者通过电话遭遇诈骗,其次是短信(50%)和即时通讯应用(40%)。购物、投资和求职诈骗最为常见。

According to the report, 56 per cent of respondents encountered scams through phone calls, followed by text messages at 50 per cent and instant messaging applications at 40 per cent.

但即使是电话诈骗也在演变,Gogolook联合创始人兼CEO郭建甫(Jeff Kuo)表示。

Shopping, investment and employment scams were the most common. But even phone scams are evolving, said Gogolook co-founder and CEO Jeff Kuo.

他警告称,AI生成的语音让电话诈骗愈发复杂,使犯罪分子能冒充公司高管甚至家庭成员。

He warned that AI-generated voices are making phone scams increasingly sophisticated, allowing criminals to impersonate company executives or even family members.

他表示,Gogolook 正利用机器学习识别可疑的来电模式,帮助其在犯罪分子频繁更换号码的情况下,依然能侦测出潜在的诈骗号码。

He said Gogolook is using machine learning to identify suspicious calling patterns, helping it detect potential scam numbers even as criminals frequently change them.

“随着诈骗分子使用 AI 生成的声音和越来越逼真的话术,在用户接触之前识别可疑号码并发出警示,变得更加重要,”他说。

“As scammers use AI-generated voices and increasingly convincing scripts, identifying suspicious numbers and warning users before they engage is becoming even more important,” he said.

GASA 报告还强调了防止重复受害的挑战,发现 51% 的受损受访者曾被骗不止一次。

The GASA report also highlighted the challenges of preventing repeat victimisation, finding that 51 per cent of respondents who lost money had been scammed more than once.

虽然 AI 能帮助当局和科技公司更高效地识别和监控可疑活动,但 Abraham 警告说,犯罪分子在采用新兴技术时,不像合法组织那样面临相同的法律和监管约束。

While AI could help authorities and technology companies identify and monitor suspicious activity more efficiently, Abraham cautioned that criminals can adopt emerging technologies without facing the same legal and regulatory constraints as legitimate organisations.

“诈骗分子不受法律或执照限制。他们能用什么就用什么。所以(虽然)我们可以用 AI 行善来识别诈骗、实现监控自动化,但这(单靠 AI)无法解决问题,”他说。

“Scammers are not limited by law or licenses. They use whatever they can get. So (while) we can use AI for good to identify scams, to automate monitoring, it (alone) will not solve the problem,” he said.

科技公司被敦促将信任置于流量之上 科技公司表示,它们正在加强针对网络诈骗的防护措施。

TECH FIRMS URGED TO PRIORITISE TRUST OVER TRAFFIC Technology firms say they are strengthening safeguards against online fraud.

Meta 安全政策经理 Rima Amin 表示,这家社交媒体巨头正在加大力度侦测和破坏诈骗网络,包括通过加强与执法机构的合作。

Meta's security policy manager Rima Amin said the social media giant is stepping up efforts to detect and disrupt scam networks, including through greater cooperation with law enforcement agencies.

“(当)我们打掉一个网络时,我们会审视诈骗分子使用的、我们的主动侦测未能捕捉到的手段。然后我们的团队吸取这些经验教训,将其构建回产品中,以便我们未来能够防御和预防,”她在峰会期间告诉 CNA。

“(When) we take down a network, we look at the tactics the scammers were using that our proactive detection didn't catch. Then our teams take those learnings and build them back into the products, so that we can defend and prevent in the future,” she told CNA during the summit.

她指出最近的一次行动,Meta 移除了 15 万个账号,而泰国警方逮捕了 21 名涉嫌参与诈骗的人员。

She pointed to a recent operation in which Meta removed 150,000 accounts while Thai police arrested 21 people suspected of involvement in scams.

但她承认,个别公司的努力是不够的,指出欺诈网站在被社交媒体平台屏蔽后,可能长期保持活跃状态。

But she acknowledged that efforts by individual firms are insufficient, noting that fraudulent websites could remain active long after being blocked on social media platforms.

Facebook和Instagram的母公司Meta发现,其诈骗拦截名单上的网站平均持续存在了三年,使其得以在其他地方继续传播。

Meta, the parent company of Facebook and Instagram, found that websites on its scam blocklist continued to exist for an average of three years, allowing them to circulate elsewhere.

“我们正在共享(我们的)拦截名单……以便其他平台能够利用这一情报保护自己。这不是任何单一平台或机构应该孤立思考的(问题)。我们需要保护整个生态系统中的人们,”Amin说。

“We're sharing (our) blocklist … so other platforms can take that intelligence and protect themselves. This isn't a (problem) that any platform or institution should think about in isolation. We need to protect people across the entire ecosystem,” said Amin.

她还强调了早期干预的重要性,即在潜在受害者确信欺诈性投资或关系是真实之前进行干预。

She also stressed the importance of intervening early, before potential victims become convinced that a fraudulent investment or relationship is genuine.

她说,Meta使用其所谓的“即时预警”在诈骗发展到财务交易阶段前,提醒用户注意可疑互动,因为那时受害者可能更难被劝阻。

She said Meta uses what it calls “just-in-time warnings” to alert users to suspicious interactions before scams reach the stage of financial transactions, when victims may be harder to dissuade.

Gogolook的Kuo表示,科技公司必须在打击诈骗方面做得更多,并警告称,数字平台往往优先考虑吸引用户,而非加强安全措施。

Gogolook's Kuo said technology companies must do more to combat scams, warning that digital platforms often prioritise attracting users over strengthening safety measures.

“我们日常使用的平台……它们最大的动力是增加流量,而不是建立更多信任机制,”他说,并补充称,数字平台之间需要更强有力的合作,并需要在旨在检测和预防欺诈活动的系统上投入更多资源。

“The everyday platforms we are using … their biggest motivation is to increase traffic instead of building more trust mechanisms,” he said, adding there is a need for stronger cooperation among digital platforms and greater investment in systems designed to detect and prevent fraudulent activity.

Kuo警告称,如果人们无法区分真实内容和欺诈材料,科技公司最终可能会失去用户。“流量可以建立平台,但只有信任才能留住平台,”他说。

Kuo warned that technology firms could ultimately lose users if people became unable to distinguish genuine content from fraudulent material. “Traffic can build a platform, but only trust can keep it," he said.

阅读媒体原文 ↗