The rise of artificial intelligence (AI) technology has brought great convenience to our daily lives, yet it also presents serious challenges. Among the most worrying concerns are deepfake scams, a new form of digital trickery.

The term “deepfake” is a mix of “deep learning” — an advanced AI method — and “fake”. This technology is trained using huge amounts of data, including pictures and recordings of human faces and voices. By studying this information, AI models can learn to perfectly copy a person’s facial expressions, body language, and tone of voice. With as little as five seconds of someone’s speaking, apps can now produce a video or sound clip that appears surprisingly real.

But the fast growth of deepfake tech has given rise to criminal activities. The old saying “seeing is believing” may no longer be true. Fraudsters now use these tools to create believable content in which people seem to say or do things that never happened.

Deepfake tricks generally fall into several groups. In investment frauds, scammers might create a video of a well-known expert supporting a false platform to steal money. Romance scams involve building false online profiles with AI-made faces to win trust and then request money. There are also emergency tricks, where a panicked call seemingly from a relative asks for immediate payment due to a sudden accident. Besides, staff imitation has become common, with fraudsters copying senior workers to get around security and reach company accounts.

Luckily, there are ways to protect ourselves from these technical tricks. During a video call, watch closely for unusual eye movements or lighting that does not seem quite right around the face. A simple test is to ask the person on screen to move a finger slowly across their face; in a deepfake, the image will often twist or become unclear. Listen carefully for any mismatch between mouth shapes and the spoken sounds, or unusual stops during speech.

1.1. What can be inferred about the AI models used to create deepfakes?

A They rely on limited samples.

B They analyze visual and audio cues.

C They alter human DNA data.

D They avoid facial expressions.

解析：选B。B推理判断题。文章第二段指出，AI模型利用包括人脸图片和声音录音在内的海量数据进行训练，通过学习这些信息来复制人的面部表情、肢体语言和语调。由此可推断，这些模型会分析视觉（图片、表情）和听觉（录音、语调）线索，B项正确。A项错在“limited samples”，原文明确提到使用“huge amounts of data”，并非依赖有限样本。C项“alter human DNA data”在文中完全没有涉及。D项“avoid facial expressions”与原文“copy a person’s facial expressions”完全相反，故排除。故选B。

2.2. What does the underlined word “Fraudsters” in Paragraph 3 most probably mean?

A People who commit crimes.

B People who design apps.

C People who report news.

D People who fix devices.

解析：选A。A词义猜测题。第三段首句提到“the fast growth of deepfake tech has given rise to criminal activities”，紧接着第二句说“Fraudsters now use these tools to create believable content in which people seem to say or do things that never happened”，可见“Fraudsters”利用这些工具伪造虚假内容，其行为属于犯罪活动。由此可推知“Fraudsters”指从事诈骗等犯罪行为的人，A项“People who commit crimes”正确。B项“设计应用的人”、C项“报道新闻的人”和D项“修理设备的人”均不符合上下文中的犯罪语境。故选A。

3.3. Which of the following is described as a reliable method to spot a deepfake during a live video call?

A Watching for eye colour changes.

B Requesting a finger-wave test.

C Demanding a written statement.

D Checking the background noise.

解析：选B。B细节理解题。根据第五段中的建议“A simple test is to ask the person on screen to move a finger slowly across their face; in a deepfake, the image will often twist or become unclear”可知，要求对方在屏幕前用手指缓慢划过脸部是一种检测深度伪造视频的可靠方法，即进行“手指摆动测试”，B项正确。原文提到观察不正常的眼球运动，而非“眼睛颜色的变化”，A项有误。C项“要求书面声明”和D项“检查背景噪音”文中均未提及，不能作为识别方法。故选B。

4.4. What is the core message conveyed in the final paragraph?

A Distrust all digital forms.

B Delete strange video apps.

C Verify through separate channels.

D Hang up on robotic voices.

解析：选C。C主旨大意题。最后一段主要提出了预防建议：对未知来电保持警惕，并强调“Before taking action, double-check the situation by calling the person back on a trusted number...”（在采取行动前，通过可信的号码回拨给当事人以再次核实情况……）。这强调了通过独立、可靠的第二渠道进行验证的重要性。选项C“Verify through separate channels.”（通过独立渠道核实）是对这一核心建议的高度概括。A项“怀疑所有数字形式”过于绝对，文中说的是“保持健康的怀疑”，并非全盘否定。B项“删除奇怪的视频应用”和D项“挂断机器人声音”均未在段落中出现，属无中生有。故选C。