Meta And TikTok Allowed More Harmful Content After Evidence Showed Outrage Drove Engagement, Whistleblowers Say

Meta and TikTok Allowed More Harmful Content After Evidence Showed Outrage Drove Engagement, Whistleblowers Say

More than a dozen whistleblowers and former employees at social media giants Meta and TikTok have told investigators that the companies made a series of decisions to prioritise user engagement over safety, allowing harmful content to be shown to users even after internal research showed that sensational, outrage‑inducing material drove interaction on their platforms.

The disclosures were made in interviews and documents provided to the BBC documentary Inside the Rage Machine, according to insiders who described how competitive pressure between the companies’ algorithms pushed them to allow more violence, harassment, hate speech, conspiracy theories and other risky content onto users’ feeds Whistleblowers said that algorithmic systems at both Meta which owns Facebook, Instagram and WhatsApp and ByteDance‑owned TikTok increasingly fed users content most likely to keep them scrolling, regardless of whether that content was inflammatory or harmful. Internal data reportedly showed that emotionally charged material such as outrage‑provoking posts generated higher engagement metrics, including comments, shares and watch time, which boost advertising revenue. A former Meta engineer, identified only as Tim, said senior management had instructed teams to allow more “borderline” harmful material to be shown on users’ feeds as the company sought to compete with TikTok’s algorithmic growth and protect its stock price. “They sort of told us that it’s because the stock price is down,” he said, according to the whistleblower testimony. Sources told investigators that safety teams at both companies were often overruled by product development teams focused on engagement and growth. At Meta, a senior researcher, Matt Motyl, said that when Instagram launched its TikTok‑style Reels feature in 2020, it did so without adequate safety safeguards even though internal research showed that comments on the feature had significantly higher levels of bullying, harassment and hate speech compared with other parts of the platform. According to whistleblower accounts, the trade‑off between growth and safety was a recurring theme: safety initiatives were delayed, under‑resourced or overridden when they conflicted with business priorities tied to user time‑on‑platform and quarterly revenue targets

Whistleblowers also described practices at TikTok in which internal dashboards of user complaints showed that political posts or relationships with public figures were prioritised over reports of content that could harm children or other vulnerable users. One insider said that staff were sometimes told to address complaints from politicians first, in part to maintain regulatory goodwill, even when reports involved serious safety concerns.

The allegations suggest that decision‑making at the company went beyond simple algorithm design and into areas where commercial and regulatory considerations influenced content moderation choices The whistleblower revelations add to growing concerns about the role of algorithmic recommendation systems in shaping online discourse and the potential risk they pose to public health, democratic processes and psychological wellbeing. Independent research and regulatory bodies have increasingly scrutinised how social media platforms balance engagement‑driven growth with safety and misinformation control a tension highlighted by the European Commission’s preliminary findings that Meta and TikTok breached certain transparency obligations under the EU’s Digital Services Act.

Meta and TikTok have not publicly confirmed all of the claims made by whistleblowers. Both companies say they invest in safety and moderation efforts, and that significant resources are dedicated to detecting and removing harmful content from their platforms. Meta’s own enforcement reports, for example, state that the company removes the majority of illegal content proactively once identified and works with external partners to improve user safety. Experts note that algorithms prioritising engagement inherently tend to amplify emotionally charged or sensational content a phenomenon known as algorithmic amplification which can inadvertently promote misinformation, hate speech and divisive material. Studies of such systems have found that content optimisation for engagement can emphasise out‑group hostility and exaggerated claims, even without coordinated manipulation, due to the way recommendation engines reward interaction. The whistleblower claims have reignited debates about social media regulation, the responsibility of tech firms to protect users (especially minors), and whether existing content moderation models are adequate in an era of powerful, AI driven recommendation engines.