Errors, on the record.

Item says the 'sole selling shareholder is Advent International' / 'only PE owner Advent International sells' — evidence shows the sole selling shareholder is AI Alpine, an entity co-owned by funds managed by Advent International AND the Abu Dhabi Investment Authority (ADIA); ADIA is omitted and the

Headline internal inconsistency: 'jumps ~15% to $33.30' conflates the opening pop (+14.81% to $31.00, which is the ~15% Bloomberg/Reuters reported intraday) with the close ($33.30 = +23.33%); a $33.30 close is a ~23% gain, not ~15%. Evidence: https://finance.yahoo.com/markets/stocks/articles/power-e

Headline says Anthropic beat 'OpenAI to the IPO line' and summary calls it 'the first major AI lab to formally start the IPO process' — but OpenAI confidentially filed its S-1 with the SEC on May 22, 2026, ten days BEFORE Anthropic's June 1 filing (the item's own context bullet even states this, con

Key claim '~80-fold annualized surge' is an arithmetic error: $4B run-rate (July 2025) to >$50B projected is ~12.5x, not ~80-fold. Fortune confirms both endpoints ($4B baseline; run-rate to 'surpass $50 billion by the end of next month') but not the multiplier. Evidence: https://fortune.com/2026/06/

Context claim 'SpaceX public S-1 May 20 (~$1.77T)' — independent coverage puts SpaceX's targeted valuation at ~$1.75 trillion, not $1.77T (consistent with prior QA flag that OpenAI/SpaceX valuations were overstated vs primary sources). Evidence: https://www.investing.com/analysis/the-trilliondollar-

Item says the April BoJ 6-3 hold had 'board members Masu and Koeda flagged as supporting a hike.' The BoJ's April 28, 2026 Statement on Monetary Policy shows Masu Kazuyuki and Koeda Junko voted WITH the 6-member majority to HOLD; the three dissenters who voted to hike were Nakagawa Junko, Takata Haj

Minor framing: 'yen near 159-160, just under perceived MoF intervention trigger' — the yen had already breached 160 (hit 160.72) in late April, which triggered the intervention; 160 was a crossed level, not merely an untouched trigger. Evidence: https://asia.nikkei.com/business/markets/currencies/ja

Item says Nightmare-Eclipse 'threatened fresh drop around July 14 (next Patch Tuesday)'. The cited Record article instead says the researcher announced a new Secure Boot vulnerability would be disclosed 'sometime in June' — not tied to July 14; neither Register article mentions July 14 either. Also,

Aliases 'Chaotic Eclipse / Dead Eclipse' for Nightmare-Eclipse appear in none of the cited sources nor independent coverage checked (The Register, The Record, Help Net Security, Picus, Barracuda); treat as unsupported. Evidence: https://www.theregister.com/security/2026/06/02/microsoft-reaches-for-o

Item says 'More than $350,000 in illicit profits (earlier reporting cited $320,000+)'. Attribution is inverted: the SEC's own civil complaint/litigation release (same day, June 5, 2026) alleges 'more than $320,000 in illicit profits'; the $350,000+ figure comes only from U.S. Attorney Jay Clayton's

Item says SEC seeks an 'officer-and-director bar'. The SEC litigation release actually says the complaint seeks 'a permanent injunction, disgorgement with prejudgment interest, civil penalties, and a conduct-based injunction prohibiting Li from acting as or associating with an investment adviser' —

Unverified detail (not refuted): assignment to U.S. District Judge Lorna G. Schofield could not be confirmed in any checked source; case number is 1:26-cv-04778 (S.D.N.Y.). Whether the criminal charge was by indictment (vs complaint) also not confirmed, though 'two counts of securities fraud' is cor

Source list error (matches prior QA flag a): the second source URL https://www.msn.com/en-us/news/insight/leaked-memo-shows-disney-plans-hulu-app-shutdown-by-2026/gm-GM11059850 is an unrelated Disney/Hulu app shutdown story, not Star Tribune coverage. Replace with relevant independent corroboration,

Item says Suleyman claimed tuned MAI models beat GPT-5.5 on enterprise tasks with a 'tenfold reduction in costs'. The primary keynote transcript attributes the tenfold cost reduction to unnamed early adopters/McKinsey examples, not to Suleyman, and no checked source (microsoft.ai transcript, TechTim

The quoted phrase trained 'entirely on clean, commercially licensed data without distillation from any third-party model' is TechTimes' paraphrase, not Microsoft's wording; the keynote says 'created with an enterprise-grade, clean and commercially licenced data lineage that you can trust' and 'train

Name misspelled 'Mustafa Suleiman' — correct spelling is Mustafa Suleyman. Evidence: https://www.geekwire.com/2026/microsoft-unveils-seven-homegrown-ai-models-in-bid-for-long-term-self-sufficiency/

Item says 'Over $100 million committed in usage credits'; Anthropic's program page says it is 'committing up to $100M in usage credits for Mythos Preview' — 'up to', not 'over'. The $4M in donations is confirmed ($2.5M to Alpha-Omega/OpenSSF via Linux Foundation, $1.5M to Apache Software Foundation)

Key claim says initial partners ('upwards of 50') were 'scanning 1,000+ open-source projects' and found the 10,000+ flaws. Per Anthropic's initial update, Anthropic ITSELF scanned the 1,000+ open-source projects (that sweep produced the 1,752 independently reviewed findings), while the ~50 partners

Minor nuance: item says 'Both NSA and Anthropic declined to comment'; the primary TechCrunch article says the NSA 'declined to confirm or deny the reporting' and Anthropic 'did not respond to a request for comment' — not a refusal to comment by Anthropic. Evidence: https://techcrunch.com/2026/06/05/

Source labeling: the URL tagged [PRIMARY] (European Financial Review, 2026-06-02) does not mention ENISA, Project Glasswing, a June 1 date, US-administration permission, or the 'several productive meetings' quote — it only says Anthropic will provide 'the European Commission' access 'after months of

Unverified detail: 'the bloc had to seek permission from the US administration before obtaining the model' was not found in any accessible source (European Financial Review, The Next Web); it is attributable only to the bot-blocked CNBC article and should carry a single-source caveat. Evidence check

The prior 40-year JGB auction (May 27, 2026) saw STRONG demand: bid-to-cover 2.70 vs 2.54 at the previous sale and above the 12-month average of 2.47 — not a weak sale; the 'third consecutive soft super-long auction' framing is wrong.

The previous 30-year auction (May 14, 2026) drew demand above its 12-month average (bid-to-cover 3.49), so June 10 was a reversal from strength, not a continuation of softness.

10-year JGB yield eased to 2.67% on June 9, 2026 (listed at ~2.68% on June 10) after oil fell ~5% on the Israel-Iran halt in attacks — substance confirmed, timing is June 9 rather than June 10.

Brent settled June 9, 2026 at $91.45/bbl, down $2.80 or 3.0% (not $89.95, down $4.30 or 4.6% — that was the 12:38 p.m. EDT intraday level)

WTI settled at $88.20, down $3.10 or 3.4% (not $86.35, down $4.95 or 5.4%)

U.S. 10-year Treasury yield was approximately 4.52% (4.54% unverified in cited sources)

Exact premium: 58.7% to Tate & Lyle's closing price on 13 May 2026 (the Undisturbed Date), per the Rule 2.7 announcement; 65.2% to the 3-month VWAP and 61.6% to the 6-month VWAP

Total value including permitted dividends is up to 615 pence per share (595p cash + up to 13.2p final dividend + up to 6.8p interim dividend), a headline premium of approximately 64.0% to the undisturbed close

Deal financed in part by a $4.225 billion 364-day bridge loan led by JPMorgan; Huber Equity Corporation gave an irrevocable undertaking covering approximately 16.8% of Tate & Lyle's issued share capital