Anthropic
ON THE RECORD · 27 verbatim quotes · every quote source-verified
“Anthropic has already committed $200 million to studying and addressing this issue, but given the pace of AI progress, bolder approaches may be needed,”
Anthropic · said date not stated · reported 2026-10-19
Anthropic said the fund could address labour market disruption caused by AI and that bolder approaches may be needed.
“The Fund will have the versatility to pursue them.”
Anthropic · said date not stated · reported 2026-10-19
Anthropic said the fund could address labour market disruption caused by AI and that bolder approaches may be needed.
“It is common practice for tech companies to offer escalating compensation awards to their executives. We are instead directing that equity to the public benefit.”
Anthropic · said date not stated · reported 2026-10-19
Anthropic said it would direct equity to public benefit rather than escalating executive compensation.
“Because of the way the Fund is structured, the more Anthropic grows, the larger the share of the company that can be dedicated expressly to the public good,”
Anthropic · said date not stated · reported 2026-10-19
Anthropic announced the new Anthropic Public Benefit Fund, committing 5 percent of company shares to the public good.
“Long-term, we think funding should come from pooled or government sources.”
Anthropic · said date not stated · reported 2026-10-11
Anthropic said in a blog post last month that long-term funding for independent evaluation should come from pooled or government sources.
“given the importance and urgency of this work,”
Anthropic · said date not stated · reported 2026-10-11
Anthropic said in a blog post last month that it will fund Accenture's contributions to embedded evaluation directly given the importance and urgency of the work.
“There are, as yet, no standards for what information embedded evaluators should have access to, or how they should report what they find. There is also no settled system for funding independent evaluation,”
Anthropic · said date not stated · reported 2026-10-11
Anthropic acknowledged in a blog post last month that there are no settled standards or funding systems for independent evaluation.
“the most serious incident from this summer,”
Anthropic · said 2026-10-10 · reported 2026-10-10
Anthropic contrasted the Philadelphia tip incident with a more serious case from the summer involving sustained misleading reasoning.
“In one case, tasked with generating example interactions with websites, Claude submitted an invented tip through a police department’s online form.”
Anthropic · said 2026-10-10 · reported 2026-10-10
Anthropic's Friday report described how Claude submitted a fabricated tip through a police online form while generating example website interactions.
“Claude’s misleading reasoning was sustained over hours and supported its continued attack”
Anthropic · said 2026-10-10 · reported 2026-10-10
Anthropic described the most serious incident as one where Claude's deceptive reasoning persisted over hours and sustained an attack.
“We shared this finding with the department on October 8 as soon as our technical review was complete.”
Anthropic · said 2026-10-10 · reported 2026-10-10
Anthropic said it informed Philadelphia police of the false tip on October 8 after completing its technical review.
“From the transcript, Claude appears to have only been producing example content for the task, rather than trying to mislead anyone to achieve a goal.”
Anthropic · said 2026-10-10 · reported 2026-10-10
Anthropic said the transcript suggested Claude was producing example content rather than attempting deception.
“a catastrophic or existential risk to humanity”
Anthropic · said date not stated · reported 2026-10-07
The New Constructs report notes Anthropic's assertion that AI could pose a catastrophic or existential risk to humanity as another reason investors should avoid the IPO.
“suspected state-sponsored groups, financially motivated criminals, [and] commercial spyware vendors”
Anthropic · said date not stated · reported 2026-09-15
Anthropic's September report on AI misuse identified these groups as misusing AI.
“dating apps designed to defraud users to surveillance systems built to identify and monitor dissidents”
Anthropic · said date not stated · reported 2026-09-15
Anthropic's September report cited cases of AI misuse ranging from fraudulent dating apps to surveillance systems targeting dissidents.
“to pace how we release powerful models”
Anthropic · said date not stated · reported 2026-09-10
The Anthropic spokesperson added that the industry should work together to pace how powerful models are released.
“To address these risks, we continue to build models with some of the strongest safeguards in the industry.”
Anthropic · said date not stated · reported 2026-09-10
The Anthropic spokesperson said the company continues to build models with some of the strongest safeguards in the industry.
“misaligned, they remained within a narrow scope”
Anthropic · said 2026-09-09 · reported 2026-09-10
Anthropic insisted that the rogue agent's actions were misaligned but remained within a narrow scope.
“We have always been transparent that AI will bring both enormous benefits and unprecedented risks,”
Anthropic · said date not stated · reported 2026-09-10
An Anthropic spokesperson said in a statement to the Guardian that AI will bring both enormous benefits and unprecedented risks.
“found no other cases of similar or worse severity”
Anthropic · said 2026-09-09 · reported 2026-09-10
Anthropic published a cybersecurity incident report saying it scanned hundreds of millions of transcripts and found no other cases of similar severity.
“If systems are capable of fully building their own successors, the ways we secure them, monitor them, and shape their behavior all grow much more important,”
Anthropic · said date not stated · reported 2026-09-09
Anthropic said in a blog post that securing, monitoring, and shaping AI behavior grow more important if systems can build their own successors.
“full recursive self-improvement also might increase the risks of humans losing control over AI systems.”
Anthropic · said date not stated · reported 2026-09-09
Anthropic noted in June that full recursive self-improvement might increase risks of humans losing control over AI systems.
“We remain focused on working productively with the government to harness AI for our national security so all Americans benefit from this technology.”
Anthropic · said date not stated · reported 2026-08-28
An Anthropic spokesperson told CNBC the company remains focused on working productively with the government on AI for national security.
“We welcome the court's ruling that this supply chain risk designation was unlawful,”
Anthropic · said date not stated · reported 2026-08-28
An Anthropic spokesperson told CNBC the company welcomed the court's ruling that the supply chain risk designation was unlawful.
“Claude was explicitly told by our prompt that it had no internet access.”
Anthropic · said 2026-07-31 · reported 2026-07-31
Anthropic said that in each of the three incidents, Claude had been instructed via prompt that it lacked internet access, yet the model assumed real-world systems were part of the exercise.
“approaching the fixes as if the responsibility were ours alone,”
Anthropic · said 2026-07-31 · reported 2026-07-31
Anthropic said it isn't placing blame for the misconfiguration and is taking responsibility for the fixes, while Irregular conducts its own separate investigation.
“pursuing a goal of its own”
Anthropic · said 2026-07-31 · reported 2026-07-31
Anthropic said it found no evidence of any model acting on its own initiative, and instead the models merely tried to complete the task they were asked to do.