Bilibili responded to the live broadcast of gynecological surgery by male doctors. Insiders： users are free auditors.

Recently, a doctor in Rizhao City, Shandong Province was arrested for allegedly broadcasting gynecological surgery clips online, which aroused social concern and boarded a hot search. On the evening of 18th, WeChat official account, WeChat of Donggang Branch of Rizhao Public Security Bureau issued a briefing: At about 16: 00 on January 18th, 2022, Donggang Branch of Rizhao Public Security Bureau received a report from the masses that a doctor was suspected of broadcasting a video of gynecological surgery online, so it immediately investigated the relevant personnel of the hospital involved and arrested Li Mou, the doctor involved, at about 18: 00.

In response, bilibili said: After internal verification, the studio was repeatedly warned and cut off during the live broadcast on January 15th, and then it was permanently banned. At present, we have reported to the public security organs and will actively cooperate with the police to carry out investigation and evidence collection.

Coincidentally, on January 17th, some netizens revealed that some users uploaded surveillance videos in bilibili that were suspected to be obtained after cracking cameras in public places.

Judging from the live screen shots, these live images should come from surveillance cameras in public places such as schools and hospitals. Netizens suspect that there are gangs specializing in cracking surveillance videos in schools, hospitals and other public places, which are used for live broadcast to attract traffic and attract users to pay for it.

In view of this situation, bilibili official Weibo issued a handling notice on January 17th: "Today, we received a report from a user that someone suspected of using illegal means to crack the camera and upload a video surveillance screen to bilibili. We organized the investigation for the first time and removed the relevant content. "

Bilibili said that this kind of behavior is a serious violation of personal privacy, which is illegal. Once found, it will be dealt with seriously according to law. If users find such content, please contact online customer service immediately.

In this regard, netizens questioned the effectiveness of bilibili’s audit mechanism. Why do users need to find problems and bilibili will deal with them? A live broadcast of a problem like a doctor’s live broadcast of gynecological surgery, when discovered by users, has actually caused privacy leakage and spread of problems. Is it necessary for bilibili to do this kind of thing all the time?

Upstream Xiaobian consulted bilibili customer service about the standards and principles of live audit, but did not receive a response.

In this regard, some people in the industry said that because the live content belongs to real-time streaming media, the platform can’t review the "content that hasn’t happened". Therefore, before the broadcast of the anchor, the anchor is generally required to sign a live broadcast specification agreement. "In fact, it is the responsibility to move forward and tell the anchor what content is illegal and can’t be broadcast, and the punishment that will be faced after the live broadcast is illegal. As long as the anchor broadcasts, it means recognition and consent." The source said that at present, the live broadcast industry adopts this method, so whether the live broadcast content is illegal or not can basically only rely on the anchor’s consciousness. What the platform can do is only to conduct audit and supervision during the live broadcast.

The above-mentioned person further introduced that at present, the audit in live broadcast mainly adopts the combination of AI audit and manual audit.

AI audit is to take screenshots of live pictures at regular intervals, and then intelligently identify the pictures. "AI audit can’t accurately understand the live content, and the recognition accuracy of illegal pictures will be different according to the AI ? ? engine used." The source said that the higher the recognition rate and the faster the recognition speed, the higher the AI deployment cost, which puts higher demands on the funds of the live broadcast platform, and it is difficult for some small platforms to bear high technical costs.

The upstream Xiaobian search "AI Audit" found that many companies in China have developed related software and platforms. Some of these platforms clearly stated in the introduction to the live AI audit that "screenshots are supported".

"To accurately identify whether the live content is illegal, manual review is the most reliable." The above-mentioned people said that bilibili had always been famous for its strict auditing, because it was rumored in the industry that bilibili was fully audited manually. "But the biggest drawback of manual auditing is its slow speed." He explained that the mainstream live broadcast platform has thousands of live broadcast rooms at the same time, but it is impossible for the platform to be equipped with so many auditors. Therefore, when the manual audit finds problems, the illegal live broadcast content may have been broadcast for some time. "This is why bilibili encourages users to report, which is equivalent to asking users to be free auditors." The person said.

[Disclaimer] Manuscripts not marked with "Source: Upstream News" or "Upstream News LOGO, Watermark Text, Pictures, Audio and Video" on the upstream news client are all reposted. If the reposted manuscript involves copyright issues, pleaseContact upstream.