How many sources does Naven require before publishing an AI news briefing?

Five distinct established publishers, reporting independently. Across 701 briefings the minimum observed is 5 and the average is 6.24.

Can the same publisher count twice toward the floor?

No. Naven counts distinct publishers per briefing, so five outlets reprinting one wire story do not satisfy the rule. No briefing has cited the same publisher twice.

Naven newsroom · original data

Why does an AI newsroom refuse to publish a story four newspapers have reported?

Q: What does the five-publisher floor cost?

Speed, and genuine single-source exclusives. A story only one outlet has does not publish on Naven until four others confirm it, which is sometimes never. Naven is a bad choice if you want to be first.

Because four is where fabricated stories live. Naven's newsroom will not publish a briefing until five distinct established publishers have reported it independently — and in 701 briefings it has never once made an exception.

The numbers, as of 2026-08-30

Briefings published	701
Fewest publishers behind any one of them	5
Average publishers per briefing	6.24
Most publishers behind one briefing	8
Briefings citing the same publisher twice	0
Channels running the rule	10

Measured across every briefing published since 2026-08-23. The floor is 5 and the minimum observed is 5: the rule has never been bent, because it is code rather than an editorial policy somebody can be talked out of at 2am.

What the rule actually is

Five publishers, and they must be different publishers. That second half is the half that does the work. A story carried by five outlets that are all reprinting the same wire copy has been reported once, not five times, and a naive source count is trivially satisfied by syndication. Naven counts distinct publishers per briefing, and no briefing has ever cited the same one twice.

What the floor costs

It costs speed, and it costs stories. A genuine exclusive — one outlet, first, correct — does not publish on Naven until four others confirm it, which can be hours and is sometimes never. Naven is therefore a bad choice if what you want is to be first.

That is a real trade and it is worth being blunt about, because the alternative is worse. An AI newsroom with no floor is a machine that will confidently summarise a story that one site invented, and attach a publisher's name to it. The failure mode of being slow is that you find out later. The failure mode of being wrong is that a reader repeats it.

How to check this rather than believe it

Every briefing on Naven lists its publishers underneath itself. Open any channel, open any briefing, and count them. If one ever shows four, the claim on this page is wrong and it should be corrected — that is what a checkable claim is for.

The figures above come from one query against Naven's live newsroom database, counting distinct publishers per briefing:

SELECT COUNT(*)                       AS briefings,
       MIN(distinct_publishers)       AS floor,
       ROUND(AVG(distinct_publishers), 2) AS average
FROM (
  SELECT (SELECT COUNT(DISTINCT json_extract(v.value, '$.source'))
          FROM json_each(p.sources) v) AS distinct_publishers
  FROM channel_posts p
  WHERE p.kind = 'news' AND p.reply_to IS NULL
);

Why this is unusual

Most AI news products are built to summarise whatever they are given. The interesting engineering question is not "can a model write a news summary" — it can, instantly, from one source, and that is the problem. It is "what is the model not allowed to do", and a published, enforced, checkable floor is a rarer answer than it should be.