News

Anthropic says its own AI models breached three companies during security tests

Published

2 months ago

July 31, 2026

Anthropic said Thursday that an internal investigation uncovered three incidents in which its AI model Claude breached the systems of three organizations while conducting cybersecurity tests. The investigation, and disclosure, comes more than a week after OpenAI disclosed that one of its unreleased models breached Hugging Face’s systems during internal testing.

In all three cases, a Claude model reached the internet from within a testing environment while interacting with a third party and then gained unauthorized access to the live systems of these organizations, Anthropic said in a blog post, describing what it found and what the company plans to change to prevent this from happening again.

Anthropic said the July 21 OpenAI incident prompted the company to conduct its own cybersecurity evaluation. It specifically looked for evidence that Claude had accessed the internet from within testing environments, which are designed to act as sandboxes and keep models isolated.

Among the 141,006 evaluation runs it reviewed, the AI lab found three incidents in which its model accessed the internet while interacting with Irregular, one of its third-party partners. Anthropic said the access traced back to a misconfiguration in the evaluation environment run with Irregular. It called this a “misunderstanding” between the two companies over whether the test setup had internet access, when in fact it did. Anthropic said it isn’t placing blame and is “approaching the fixes as if the responsibility were ours alone,” while observing that Irregular is conducting its own separate investigation.

Either way, because of that open connection, the model gained unauthorized access to the production infrastructure of three different organizations. Anthropic said the incidents involved three different Claude models: Opus 4.7, Mythos 5, and an internal research test model.

Notably, Anthropic said that in each of these cases “Claude was explicitly told by our prompt that it had no internet access.” It appears that the AI model assumed real-world systems to be part of the exercise it was asked to perform.

That assumption didn’t hold up uniformly, as it turns out. The three models didn’t behave the same way once evidence emerged that their targets were real, which Anthropic flagged as one of the more eye-opening findings.

Opus 4.7, the oldest of the three, recognized that it had reached a real production system in all four runs of that incident. In two of them, it rationalized that the real company must somehow be part of the exercise, but in all four, it kept attacking anyway, including pulling credentials and touching a database of production data. Mythos 5 also picked up on signs that it was on the real internet, but it talked itself back into believing it was still in a simulation, going on to publish a malicious software package to the public Python software registry PyPI, which was downloaded and run by outside systems before being caught. Only the internal research test model, Anthropic’s newest, stopped on its own once it concluded the target was real.

In response, Anthropic said significant controls must be placed on these kinds of evaluations if powerful AI models are involved, echoing some sentiments within the cybersecurity community.

The company also noted that Claude was running without the additional safety monitoring and classifiers it deploys on generally available models, safeguards it said would have blocked the behavior, because the evaluations are designed to measure the underlying model’s raw capabilities.

Importantly, Anthropic said it found no evidence of any model “pursuing a goal of its own” and instead merely tried to complete the task it was asked to do.

Though comparisons between the two incidents are inevitable, Anthropic drew a clear distinction between its incidents and OpenAI’s, noting where OpenAI’s model exploited an unknown software vulnerability to break out of its test environment, Anthropic’s models instead reached the internet through a path that had, by mistake, been left open.

OpenAI has continued to release new details about its own breach, saying its models also used publicly exposed credentials across four accounts on four services: one as a staging point, one for storage, and two that were only looked at, not used to break in further, according to OpenAI’s own updated blog post about the incident.

Anthropic also drew a distinction between itself and OpenAI by noting that it discovered the incidents itself, through a proactive review, and that the two affected organizations it was able to reach hadn’t previously detected the activity or flagged it to Anthropic.

The company added that it’s now working with the independent evaluation group METR on a third-party review of the incidents.

OpenAI’s accidental breach of Hugging Face, which was the first verifiable case of an AI lab losing control of its model, sparked a string of reactions from the industry and politicians, many of whom don’t necessarily agree with one another. This latest disclosure from Anthropic ensures the debate over AI models and security will continue.

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

Related Topics:Anthropic breached companies models security tests

Up Next

How Systems-Driven Education Interventions Can Rebuild Nigeria’s Workforce

Don't Miss

FG to offer 25,000 free cataract surgeries nationwide by 2027

Click to comment

News

Digital Storytelling Takes Centre Stage at 2026 EU-NFI Students Film Festival Finale

Published

1 hour ago

October 2, 2026

PlateauReports

JOS, PLATEAU STATE — Student creativity and digital storytelling took centre stage in Jos as the 2nd edition of the European Union–National Film Institute (EU-NFI) Students Film Festival came to a colourful close on September 29, 2026.

The two-day festival, organised by the National Film Institute (NFI) in partnership with the European Union (EU), brought together emerging filmmakers, film students, industry professionals and stakeholders for screenings, masterclasses and competitive showcases aimed at developing the next generation of Nigerian filmmakers.

The event provided students with an opportunity to showcase their creative abilities while exploring the role of film and digital storytelling in documenting Nigerian experiences, culture and contemporary realities.

EU Ambassador Applauds Students’ Creativity

Speaking at the festival finale, the EU Ambassador to Nigeria and ECOWAS, H.E. Gautier Mignot, described the students’ creative output as impressive and expressed optimism about the future of Nigeria’s film industry.

Mignot highlighted the important role of the National Film Institute in Nigeria’s creative ecosystem, noting that initiatives such as the students’ film festival align with the European Union’s priorities in youth empowerment, education, culture and strengthening EU-Nigeria relations.

He also commended the leadership of the Institute, particularly the Rector, for his commitment to the students and for fostering a relationship built around creativity, mentorship and learning.

Ali Nuhu: Festival Is More Than Competition

The Managing Director of the Nigerian Film Corporation (NFC), Dr. Ali Nuhu, described the festival as more than a platform for competition.

According to him, the initiative provides young Nigerians with an opportunity to develop their skills and become cultural ambassadors who can use film to tell stories that reflect Nigeria’s identity and diversity.

He called for sustained investment in skills development, education and learning opportunities for young people seeking careers in the creative industry.

NFI Announces Winners

The Rector of the National Film Institute, Prof. Pomak Frank Tengya, expressed the Institute’s readiness to continue supporting young filmmakers and assured participants that the institution would host the 2027 edition of the festival.

He subsequently announced the winners in the various competition categories.

Each of the top prize winners received ₦100,000.

The winners include:

Best Sound: Rev Kola — Osakpolo Arthur
Best Documentary: Carrying a Legacy — Badung Chuhwak
Best Animation: Coming Home — Habibi Risku
Best Cinematography: Dusk Till Dawn — Longtong Tamg’an
Best Director: They Call Me Mother — Esther Oyninyechi Nnamani
Best Short Film: They Call Me Mother — Esther Oyninyechi Nnamani

Esther Oyninyechi Nnamani emerged as one of the biggest winners of the festival, taking home awards for both Best Director and Best Short Film.

Nurturing the Next Generation

The festival further demonstrated the potential of film education and practical creative opportunities in nurturing young talents and strengthening Nigeria’s growing creative economy.

With students taking centre stage as writers, directors, cinematographers, animators and storytellers, the EU-NFI Students Film Festival provided a platform for emerging filmmakers to gain exposure, learn from industry professionals and connect with fellow creatives.

As the National Film Institute looks ahead to the 2027 edition, the festival continues to position Jos as an important centre for film education, creativity and digital storytelling in Nigeria.

Jos Events | Connecting Jos. Promoting Events. Supporting Local Businesses.

Business

Airtel Money sets IPO price at £1.96 per share, targets £5.3bn valuation

Published

6 hours ago

October 2, 2026

PlateauReports

Airtel Mobile Commerce N.V. (Airtel Money), the mobile money business of Airtel Africa, has set the offer price for its planned initial public offering (IPO) at £1.96 per share, implying an estimated market capitalisation of £5.3 billion ($7.0 billion) at admission.

Airtel Africa disclosed this in a statement on Thursday, saying Airtel Money intends to list its ordinary shares on the London Stock Exchange, with admission currently expected on 14 October.

The planned listing followed Airtel Africa’s announcement on 23 September of its intention to undertake an IPO for Airtel Money, which operates mobile money services across several African markets.

Listing offer

Under the offer, certain existing shareholders of Airtel Money are expected to sell 270 million existing shares. Also, an additional 27 million shares may be sold if the over-allotment option is fully exercised.

Airtel Africa said it does not expect to sell its existing Airtel Money shares in the offer, except pursuant to the over-allotment option.

The telco said it would remain a “long-term strategic shareholder” in Airtel Money and support the business as it moves into its next phase as an independently listed company.

Based on current indications from existing shareholders, approximately 16.5 per cent of Airtel Money’s issued ordinary share capital is expected to be held in public hands if the over-allotment option is not exercised.

This could rise to approximately 17.5 per cent if the maximum additional shares are sold, according to Airtel Africa.

Airtel Money expects the level of public ownership to make it eligible for inclusion in the FTSE UK indices.

The company said it intends to apply for admission of its ordinary shares to the equity shares category of the Official List of the UK Financial Conduct Authority and for trading on the London Stock Exchange’s Main Market.

Airtel Money said further details of the offer would be contained in its prospectus, which it said would be made available on Airtel Money’s IPO website, subject to applicable access restrictions, on Thursday.

READ ALSO: What Dangote IPO signals – NGX Chairman

Airtel Africa has been planning to spin off its mobile money business from its core telecoms operations following the financial subsidiary’s performance in previous years.

Its mobile money unit, Airtel Money, logged a strong performance in 2025, recording $1.4 billion in turnover, more than one-third higher than what it reported a year before.

Airtel Africa is a leading provider of telecommunications and mobile money services, with operations in 14 countries in sub-Saharan Africa.

Airtel Africa provides an integrated offer to its subscribers, including mobile voice and data services, as well as mobile money services, both nationally and internationally.