Product1 distinct publisher2 min readUpdated
After seven deaths on the A66, Luke Pollard told platforms to demote dangerous driving as well as delete it. TikTok answered with a takedown rate that measures something else.
The Product Desk · Product desk

Compiled by The Product DeskSomething wrong?How this is made
Follow any of these and your For You feed starts watching them — no settings page required.
The two verbs in Pollard's sentence sit on different legal footings [2]. Removal is the one Downing Street pointed at, saying the culture secretary has been clear that footage should come down "in line with their legal obligations", with Ofcom holding the government's full backing to act against platforms that fail in their duties [8]. Ranking is the one he added, and the Act as described by the BBC does not put ranking anywhere: video featuring illegal activity is not necessarily illegal itself, and it falls to the platform to decide whether flagged content is unlawful and act accordingly [9]. The carve-outs are harder than vague. Footage of a person being killed is not automatically illegal unless a terrorist group is sharing it, or children can reach it [10]. A platform can therefore satisfy the removal duty in full and go on recommending the genre.
That is close to what TikTok's own answer describes, without intending to. The company says its policies prohibit content that could cause significant physical harm, including dangerous driving, and that between January and March 2026 it removed 99.8% of content taken down under those policies proactively [12]. The denominator there is removals, not uploads or impressions, so the number reports on its own moderation queue rather than on what was served to anyone. The residual works out at one removal in every five hundred that waited for someone else to flag it [19]. TikTok also accepts that a takedown is not a finding of illegality, and may reflect nothing more than its own terms and conditions [13].
The material BBC Verify examined shows why "not celebrated" is a ranking question rather than a legal one. A 94-second clip posted to the account in March 2025 showed a car being driven across footpaths and parkland in Ormesby, Middlesbrough, with Saddington in a rear seat [14]. Another, since deleted, appears to show a joyride with the speedometer reading 100mph, and later a different car set on fire [15]. Taken one at a time, each is an upload awaiting a judgement call. Taken together they are a format, and formats are what a recommender is for. Pollard's other line was the operative one: the companies know exactly what content is on their sites, and they know how their algorithms work [6].
Meta and X were contacted for comment and no response from either appears in the BBC's report [16]. That gap matters more than it looks, because TikTok at least has a figure to put on the table, even though the figure answers a question nobody in government asked.
Ranked by verification strength, evidence, and original report placement.
Defence Minister Luke Pollard called on social media companies to remove posts celebrating dangerous driving, after a viral trend showing people going the wrong way down roads.
Pollard told BBC Breakfast on Monday: "Social media companies could take that down today, adjust their algorithms so it's not celebrated."
On Saturday PC Tom Clough, 38, and PC Matthew Blades, 37, died when a Volkswagen Passat, which officers had been pursuing until seconds before, collided with their marked armed response car on the A66 near Middlesbrough; all five occupants of the Passat also died.
A TikTok account believed to belong to one of the five passengers, 18-year-old Makai Saddington, shared videos appearing to show dangerous driving, which have since been removed.
Pollard said of the platforms: "They know exactly what content is on their site, they know how their algorithm works."
Pollard said people post content to get views and that more views sometimes bring money, so there is a financial incentive to post outrageous content that gets more people watching it.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One outlet, but named officials, on-record platform statement and first-party clip analysis
All material comes from a single publisher, which caps corroboration. Within it, the evidence is unusually concrete: named victims and officers, direct quotes from a minister, Downing Street, Ofcom and the Cleveland PCC, an on-record TikTok statement, and BBC Verify's own frame analysis of a 94-second clip locating it in Ormesby. What remains unverified is the load-bearing quantitative claim: TikTok's 99.8% proactive figure is self-reported with no denominator or audit, and no legal citation is given for a duty covering recommendation changes.
Deletions and an open probe; no evidence of ranking change
Concrete uptake is limited to enforcement on one account, a self-reported quarterly removal ratio, and Ofcom's July investigation into TikTok's child safety duties. The specific action ministers asked for - demoting or de-celebrating this content in recommendation systems - has no reported instance from any platform, and Meta and X did not respond. Removal after a fatal crash of a clip that had been live since March 2025 is weak evidence that existing detection reaches this category at speed.
Demand and metric both overstate what is enforceable or measured
Positive gap on two counts. Ministers and Downing Street frame the ask in the language of legal obligation and Ofcom enforcement, yet the statutory description supplied leaves the illegal/legal determination to platforms, carries exceptions the report calls hard to interpret, and contains no stated duty to alter ranking - so the 'adjust their algorithms' demand outruns its cited legal basis. TikTok's reply, in turn, answers a distribution question with a detection statistic. The gap is moderate rather than extreme because the underlying harm is documented with seven deaths, verified clips and a comparable Kildare crash.
Political, platform and creator incentives all explicit in the record
The incentive structure is unusually visible. Pollard himself names creator monetisation as a driver of outrageous posting. The government has an evident interest in demonstrating action days after two officers died, which is served by directing demands at platforms and by invoking Ofcom's backing. TikTok has an interest in publishing its most favourable moderation statistic - a near-perfect proactive rate - rather than time-to-detection or reach figures. Meta and X, silent here, face no comparable disclosure pressure in this report.
Facts of the incident firm; enforcement reading rests on one outlet
High confidence in the incident record, the quotes and the existence of the deleted videos, all of which are on-record or first-party verified. Lower confidence in the analytical core - that no statutory line exists for the algorithm change requested - because it rests on a single outlet's legal characterisation with no provision cited and no regulator confirmation on that specific point. Confidence in the platform-effectiveness picture is likewise limited by unaudited, self-reported figures and two absent platforms.
product
Cinemas, classrooms and ICE: smart glasses now need a venue-policy contingency1 distinct publisher
product
Senators want every TikTok test that disabled a safety feature, not just the one that hit 15 million3 distinct publishers
invest
X's stablecoin creator payouts would be a plumbing decision, not a crypto bet3 distinct publishers
product
Australia just made refusing to negotiate a taxable act, not a product decision1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 24, 2026