← All passages
Key idea: Reading to learn

An informational text explains a topic with facts, examples and explanations. Each paragraph usually has a topic sentence that the rest supports.

Key words

Main idea
What the whole text is mostly about, not just one part.
Detail
A fact the text says directly. Go back and find the sentence.
Word meaning
Use the words around a word to work out what it means here.
Inference
Something the text doesn't say, but the clues show. Clue + what you know = inference.

🔎 What to look for: Read the first sentence of each paragraph to map the text, then go back to the right paragraph for details.

Seeing Is No Longer Believing

Grade 11 · Article · 973 words
Speed

For most of the twentieth century, a photograph or video was treated as strong evidence that something had really happened. Courts accepted photographs as proof. News organizations built their reputations on footage from the scene. A famous image could change public opinion overnight. Images could be edited, of course: dictators had rivals airbrushed out of official photographs, and darkroom technicians could combine negatives. But convincing fakes took skill, time and expensive equipment, and experts could often detect them. Artificial intelligence has changed that.

"Deepfakes" are images, audio recordings or videos created or altered by AI to show people saying or doing things they never said or did. The name combines "deep learning," a type of machine learning that uses layered networks loosely inspired by the brain, with "fake." The technology learns from large amounts of real footage of a person, then generates new material that copies their face, voice and mannerisms. Early deepfakes, around 2017, were often easy to spot: faces flickered at the edges, eyes did not blink naturally, and lighting did not match. Each year, these flaws have become harder to see. Today, free or inexpensive apps can produce results that fool many viewers, and a convincing copy of someone's voice can sometimes be made from less than a minute of recorded speech.

The technology is not inherently harmful. Filmmakers use it to de-age actors or dub movies into other languages with matching lip movements, so that an actor appears to speak fluent Japanese or Portuguese. People who have lost their voice to illness, such as amyotrophic lateral sclerosis (ALS), have used AI voice cloning to speak again in something close to their own voice, built from recordings made before their illness progressed. Museums have used it to let visitors "interview" historical figures. Teachers have used AI-generated characters to make history and language lessons more engaging.

The risks, however, are serious. Deepfakes have been used to spread false political messages, such as fake recordings of candidates making shocking statements days before an election, when there is little time to correct them. They have been used to create fake celebrity endorsements in online scams, in which a well-known person appears to recommend an investment or a product. They have been used to harass people, particularly women, with fabricated intimate images, causing deep and lasting harm to victims who never did anything wrong. Criminals have used cloned voices to impersonate family members in phone calls asking for money, sometimes pretending that a grandchild has been arrested or injured. In one widely reported case, an employee of a multinational company was tricked into transferring millions of dollars after a video call in which every other participant, including the company's chief financial officer, was a deepfake.

Perhaps most troubling is what some researchers call the "liar's dividend": when fakes become common, real evidence can be dismissed as fake, too. A politician caught on a genuine recording can simply claim it was generated by AI, and some people will believe them. In this way, deepfakes threaten not only to make people believe false things, but to make them doubt true things. A society in which nobody trusts any recording is a society in which it becomes much harder to hold powerful people accountable.

Responses are developing on several fronts. Technology companies are working on tools that detect manipulated media, though detection is a constant race, since each improvement in detection can be used to train better fakes. Others are developing "content credentials" that attach information about how an image was made, such as which camera captured it and whether it was edited, in a way designed to be difficult to remove. Some AI companies add invisible watermarks to the images their systems create. Several governments are introducing laws targeting harmful deepfakes, especially non-consensual intimate images and election interference, and some platforms now require labels on AI-generated content.

But many experts believe the most important defence is media literacy: checking the source, looking for the same story from reliable outlets, and pausing before sharing something that provokes a strong emotional reaction. A sensational clip that appears only on one account, without any reliable news organization reporting it, deserves suspicion. Families are also being encouraged to agree on a private code word that a relative can use to prove their identity if they ever call asking for urgent help.

Some of the most common harms are not political at all. Criminals have used cloned voices, built from just a few seconds of audio taken from social media, to phone parents or grandparents and pretend to be a relative in trouble who urgently needs money. In 2024, voters in the American state of New Hampshire received robocalls in which an AI-generated voice imitating President Joe Biden told them not to vote in a primary election; the person behind the calls was later fined millions of dollars. Students have also been targeted by classmates who used apps to create humiliating fake images, a form of bullying that can spread through a school in hours and is extremely hard to erase. In Canada and many other countries, sharing certain kinds of fake intimate images is now a crime.

There are also small habits anyone can practise. Be most suspicious of content that makes you feel a strong emotion, such as outrage, fear or delight, because that is exactly what manipulators aim for. Pause before sharing. Check whether the clip appears on the account or website of the person it claims to show. For a phone call from a "relative" asking for money, hang up and call them back on a number you already know; some families agree on a secret word for emergencies. None of these steps is perfect, but together they make deception much harder.

In a world where seeing is no longer believing, thinking carefully has never mattered more.

Questions

Main idea 1. What is the central idea of the article?
Detail 2. Which is a beneficial use of the technology mentioned in the article?
Word meaning 3. What is the "liar's dividend"?
Reading between the lines 4. Why does the author advise pausing before sharing something that provokes a strong emotional reaction?
Reading between the lines 5. How is the article organized?
Word meaning 6. Where does the word "deepfake" come from, according to the article?
Reading between the lines 7. Why does the article say detecting deepfakes is "a constant race"?
Detail 8. What do experts suggest families agree on to protect against voice-cloning scams?
Zero2Hero — Zéro2Héros