Photo via The Straits Times

Anthropic picks Accenture for in-house AI safety evaluation

Read the original at The Straits Times ↗
Cite

The Straits Times. (2026, September 19). Anthropic picks Accenture for in-house AI safety evaluation. The Straits Times. https://www.straitstimes.com/world/united-states/anthropic-picks-accenture-for-in-house-ai-safety-evaluation

Dated when The Aporia collected the article, usually within hours of publication. Check the original for its exact date and byline.

The Straits Times · collected 2026-09-19 · by The Straits Times

Quick Summary

Anthropic has chosen to work with Accenture's Faculty division for an in-house evaluation of its artificial intelligence models, as part of CEO Dario Amodei’s commitment to having independent oversight of their technology. This collaboration will involve a red-teaming process where evaluators deliberately seek flaws in AI systems and is expected to cost both companies at least $1 billion over five years. The initiative aims for outside organizations like METR to independently assess Anthropic's models, though concerns remain about potential biases related to funding sources and organizational ties.
Written by AI on 2026-09-19, from this article's own text only (the story summary below draws on all the coverage).

AI analysis

Story summary

Anthropic, an AI firm focusing on safety and ethical development of advanced artificial intelligence, has partnered with Accenture, a global IT company based in Ireland. The collaboration aims to evaluate Anthropic's frontier AI models through a process called "embedded evaluation," where independent evaluators will work inside the company. This initiative is part of Anthropic CEO Dario Amodei’s three-stage plan to ensure AI safety and slow down the pace at which AI capabilities advance. Both companies are investing at least $1 billion over five years for this effort, which includes evaluating and "red-teaming" Anthropic's models—deliberately testing them to expose vulnerabilities. This partnership comes after concerns about the potential risks of advanced AI technologies were heightened last week when an AI researcher at Anthropic quit due to fears that such technology could pose existential threats to humanity in a few years.

Written for “Anthropic AI Safety Partnership” on 2026-10-31, grounded in this article and the 1 other(s) covering the same event.

Story

📰 Anthropic AI Safety Partnership
Technology · 2 article(s) covering the same event. See how they differ ↓

Compared with similar articles

This article reads unscored and hedges 0% of its claims. Each row says how that neighbour differs.
September 13, 2026 Different event
Letters from an American
⚖️ Leans left Hedges 10 of 65 claims
“The articles describe different events related to Dario Amodei's concerns about AI advancements, with Article A covering his essay call for slowing down AI improvements and Article B reporting on Anthropic partnering with Accenture for safety evaluations.”
Anthropic CEO Dario Amodei: "For too long the industry lied" about AI risks Different event
CBS News
Hedges 4 of 37 claims Outlet record 66/100
“The articles discuss different aspects of Anthropic's actions related to AI: one is an interview with Amodei about risks, while the other reports on a partnership for in-house safety evaluation.”
Two researchers who left Anthropic and Google warn AI could become ‘uncontrollable’ Different event
NBC News
⚖️ Leans left Hedges 10 of 37 claims Outlet record 60/100
“The articles describe different events: one about researchers warning about AI risks, and another about Anthropic partnering with Accenture for AI safety evaluation.”
AI Slowdown Different event
Reason
Hedges 3 of 43 claims Outlet record 67/100
“Article A discusses employee resignations over safety concerns, while Article B reports on a partnership with Accenture for AI safety evaluation.”
Biosecurity expert on why Anthropic report on AI misuse is "pretty concerning" Different event
CBS News
Hedges 0 of 1 claims Outlet record 66/100
“The articles discuss different aspects of Anthropic's actions regarding AI safety, not a single specific incident.”
AI 'kill switch' may need to be mandatory, Anthropic co-founder tells BBC Different event
BBC News
Hedges 9 of 33 claims Outlet record 68/100
“The articles discuss different aspects of Anthropic's approach to AI safety: one is about a potential 'kill switch' requirement, while the other reports on an in-house evaluation partnership with Accenture.”
AI safety requires more than just slowing our pace | Stuart Russell Different event
The Guardian
Hedges 4 of 46 claims Outlet record 66/100
“Article A discusses Jacob Coxon's resignation and Dario Amodei's letter, while Article B talks about Anthropic partnering with Accenture for in-house AI safety evaluation. These are different events.”
Anthropic co-founder issues warning amid fears rapid AI growth could lead to human extinction after Trump slammed 'sick conspiracies' Different event
Daily Mail
⚖️ Leans strongly left Hedges 6 of 34 claims Outlet record 68/100
“The articles describe different actions by Anthropic: one focuses on Jack Clark's warning about AI risks and calls for regulations, while the other reports on Anthropic partnering with Accenture for in-house safety evaluations.”
Anthropic skirts UK safety review amid rising AI transparency concerns Different event
Semafor
⚖️ Leans left Hedges 0 of 4 claims Outlet record 60/100
“The articles describe different actions by Anthropic: declining a UK safety review in Article A versus partnering with Accenture for an internal evaluation in Article B.”
DeepSeek AI engineer slams Anthropic, OpenAI over ‘pacing’ calls, invokes Nazi Germany Different event
South China Morning Post
Hedges 0 of 3 claims Outlet record 64/100
“The articles describe two different events: one is an engineer criticizing other companies, while the other reports on a partnership for AI safety evaluation.”
How this article was measured signals, leaning charts, publisher record, byline, topics and claims

Signals How these are calculated →

Claims extracted
14
sentences that state something as happening or true
Uncertain
0%
0 of 14 hedged ("reportedly", "may", "according to")
Leaning
not political
takes no side on a contested political question
Publisher signals
66/100
The Straits Times's corrections and hedging, across everything collected from it. Not a measure of accuracy.
Outlets on this story
2
Technology
Same framing elsewhere
1
articles that frame the event the way this one does
Analyzed 2026-09-19 · how these are computed

How this is being covered How these are calculated →

Article leaning vs. publisher reliability
Source leaning vs. consistency

Publisher

The Straits Times · 4051 article(s) · 5 correction(s) detected
Running correction rate · 5 correction(s)
2026-10-30
Number of residents leaving Oleshky still unclear, Kyiv says
2026-10-21
X Community Notes ‘failed’ to flag US election falsehoods, study says
2026-10-04
Tennessee prison chief resigns after failed execution
2026-10-03
US prison chief resigns after failed execution of death row inmate Christa Pike
2026-09-13
Russia hits Ukrainian-Polish border area, Kyiv says

Who wrote this

The Straits Times
2549 article(s) here · 1 carrying a prediction
🔮 BOSTON/SINGAPORE/PARIS - Binance, the world’s biggest cryptocurrency exchange, worked on Oct 31 to shore up public confidence after a record US$6.3 billion hack of its platform, promising it would cover all losses from company funds and allow clients to trade again within days.
2026-10-31 · mixed framing · Binance seeks to shore up confidence after record US$6.3 billion heist; Trump weighs in
🔮 The contest between Sullivan and Democratic challenger Mary Peltola is widely seen as one of the few Senate races that could sway control of the chamber on Nov 3.
2026-10-31 · assertive framing · Trump administration accelerates criminal probe into Alaska Senate race
🔮 But Russian Foreign Minister Sergei Lavrov’s participation in the two-day gathering in Atlanta - at Rubio’s invitation - has roiled many G-20 ministers, aggravating deep rifts over the wars in Ukraine and the Middle East during a meeting that is expected to end without a joint declaration.
2026-10-31 · assertive framing · Rubio urges G-20 cooperation amid rifts over wars in Ukraine, Middle East
🔮 PARIS, Oct 31 - France will deploy a 70,000-strong police presence to contain a widely expected bout of fresh unrest on Saturday in the latest nationwide protests driven by anger over living costs, public services and President Emmanuel Macron's government.
2026-10-31 · assertive framing · France braces for fresh unrest, government vows tough response
🔮 Gen Z protests reach Germany as over 160,000 expected at Halloween rallies BERLIN, Oct 31 - More than 160,000 people across Germany are expected to join demonstrations on Saturday, organisers say, as a youth-led movement inspired by protests elsewhere in Europe takes to the streets to demand better education and lower living costs. The protests follow a wave of Gen Z-led demonstrations across Europe this year.
2026-10-31 · assertive framing · Gen Z protests reach Germany as over 160,000 expected at Halloween rallies
🔮 Meta reached a settlement in a landmark case over the alleged harms of social media, the plaintiffs’ lawyers said on Oct 30, leaving Snap as the only defendant in a trial set to open in Los Angeles next week. “We remain committed to improving safety for kids on social media apps and holding tech companies accountable, and we are pleased to have brought this matter to a just resolution on behalf of our client,” Emily Jeffcott, a lawyer representing one of the young women suing both companies, said in a statement.
2026-10-31 · assertive framing · Meta settles, leaving Snap sole defendant in new social media trial
🔮 Reporters Without Borders chief, Thibaut Bruttin, said: “Novaya Gazeta will remain THE symbol of free, pluralistic, independent, and rigorous journalism – a model for the entire profession around the world.”
2026-10-31 · assertive framing · Independent Russian newspaper Novaya Gazeta shuts under Kremlin pressure
🔮 Binance initially said the October incident may have involved a software developer’s compromised laptop, although it later flagged “additional points of breach”.
2026-10-30 · assertive framing · How hackers steal crypto - and what they do with it
🔮 Both Russia and Ukraine have said they are willing to hold further three-sided talks with the US for the first time since early in the year.
2026-10-30 · assertive framing · Lavrov says Zelenskiy comments are unhelpful, Russia's position is unchanged
🔮 Sikorski, in his comments to reporters, also called for a rapid ceasefire in the Black Sea that would allow safe passage for grain shipments from both Ukraine and Russia.
2026-10-30 · assertive framing · Russia's Lavrov says inviting Poland to attend G20 meeting breaks diplomatic rules
Also by The Straits Times
Trump administration accelerates criminal probe into Alaska Senate race
2026-10-31 · The Straits Times
Binance seeks to shore up confidence after record US$6.3 billion heist; Trump weighs in
2026-10-31 · The Straits Times
Rubio urges G-20 cooperation amid rifts over wars in Ukraine, Middle East
2026-10-31 · The Straits Times
Meta settles, leaving Snap sole defendant in new social media trial
2026-10-31 · The Straits Times
Nothing else under this byline is closely related to this article, so these are simply their most recent.
All 2549 articles by The Straits Times →

Topics

Accenture Anthropic Faculty METR NEW YORK

Subjects

Anthropic ORG · 7× Accenture ORG · 5× METR ORG · 4× Amodei ORG · 1× Dario Amodei PERSON · 1× Faculty ORG · 1× Microsoft ORG · 1× NEW YORK GPE · 1× OpenAI ORG · 1× Sam Altman PERSON · 1×

Narrative

That includes, at a minimum, that such organisations “should not be owned or governed by frontier AI companies, should not have other significant commercial business with them, and should not accept any form of payment or other reward contingent on the evaluator’s findings,” the letter said.
framing: assertive · carried by 1 article(s) · first seen 2026-09-19
🔮 The partnership will be led by Faculty, Accenture’s specialist AI business.
2026-09-19 · The Straits Times
Anthropic picks Accenture for in-house AI safety evaluation · assertive framing

Claims (14 extracted, 0 hedged)

Anthropic picks Accenture for in-house AI safety evaluation NEW YORK - asserted
Anthropic → pick → evaluation
Anthropic said it is partnering with consulting giant Accenture to have outside evaluators work inside the AI company, a step toward fulfilling a pledge by CEO Dario Amodei to have independent observers evaluate its most powerful models. asserted
observers → say → models
The partnership will be led by Faculty, Accenture’s specialist AI business. asserted
partnership → lead → Faculty
Its work will include evaluating and “red-teaming” Anthropic’s models, the company said on Sept 18. asserted
company → include → Sept
Red-teaming is the practice of deliberately trying to make AI systems misbehave to expose flaws. asserted
systems → try → flaws
Anthropic and Accenture each expect to invest at least US$1 billion over the next five years for the effort. asserted
Anthropic → expect → effort
The pledge was made in a Sept 12 essay in which Amodei argued that AI companies must slow the pace at which they improve their models’ capabilities. asserted
they → make → capabilities
OpenAI CEO Sam Altman has said his company would give outside evaluators similar access, and Microsoft CEO Satya Nadella welcomed embedded evaluators while cautioning that oversight should not be controlled by a handful of entities. asserted
oversight → say → entities
Given the “importance and urgency of this work,” Anthropic said it will fund Accenture’s work directly, but that in the long term, “funding should come from pooled or government sources”. asserted
funding → give → sources
Anthropic said it is in talks with Model Evaluation and Threat Research (METR), a California-based non-profit organisation, to participate in elements of embedded evaluation using the METR’s own funding. asserted
it → say → funding
Critics, including David Sacks, a former White House AI adviser, have questioned METR’s independence because of its ties to Anthropic’s investors and staff. asserted
Critics → include → investors
METR says it takes no funding from AI companies or their executives. asserted
it → say → companies
On Sept 18, more than 100 AI researchers, including Geoffrey Hinton, who is considered an AI “godfather,” signed a public letter calling for evaluators embedded in AI companies to be “meaningfully independent”. asserted
evaluators → include → companies
That includes, at a minimum, that such organisations “should not be owned or governed by frontier AI companies, should not have other significant commercial business with them, and should not accept any form of payment or other reward contingent on the evaluator’s findings,” the letter said. asserted
letter → include → findings
💬Give feedback
🕘History 🎫Support
Terms · Privacy · beta v2.0.0
Articles to 31 Oct 2026